# PMA — Prediction Market Arbitrage Resource Hub (full text) ## What this site is PMA is an independent resource hub for prediction market arbitrage. It lists scanners, bots, guides, code libraries, data/math tools and services, each with a short description and a direct link. It is not affiliated with the listed sites unless stated. Not financial advice; capital is at risk. ## What prediction market arbitrage is The same event is often priced differently on two venues (for example Polymarket and Kalshi). If buying YES on one venue and NO on the other costs less than $1.00 combined, the position pays $1.00 at resolution regardless of outcome. Real edge depends on fees, slippage, liquidity, capital lock-up time and resolution risk (venues may resolve the "same" question differently). ## Resource directory Scanners: PredictionMarketsPicks (https://predictionmarketspicks.com/), ArbBets (https://getarbitragebets.com/), StartupHub.ai live arb feed (https://www.startuphub.ai/arbitrage). Bots: Neural Bet (independent partner guide at https://predictionmarketarbitrage.com/neuralbet), POLBOTS (https://polbots.com/), AgentBets.ai (https://agentbets.ai/). Guides: Predictefy (https://blog.predictefy.com/), PredictPedia (https://predictpedia.com/), Novig Arbitrage Guide (https://novig.com/blog/prediction-market-arbitrage-guide-2026). Data & Math: ArbLedger (https://arbledger.com/), PredictionNerd (https://predictionnerd.com/). Code: Polymarket CLOB client (https://github.com/Polymarket/py-clob-client), Kalshi API docs (https://docs.kalshi.com). Services: LaunchPoly (https://launchpoly.com/), Polymarket (https://polymarket.com), Kalshi (https://kalshi.com). More scanners: Oddpool (free / $30 Pro), prediction.com (free + Pro WebSocket), Apify actors (~$0.05/run). ArbBets (from $59/mo) figures are self-reported. More code: pmxt (TS/Python SDK, 16 venues, dry-run arb bot), Synpath (Python SDK, smart order routing), official Polymarket CLOB clients (npm/Rust), Kalshi REST/WebSocket with demo. Data: Polymarket public CLOB (no key), Jon Becker dataset (Hugging Face), PMXT Archive, Synpath Kalshi history, PredictionData.dev, CryptoStruct. Kalshi restricts commercial redistribution of its data. Venues: Polymarket US, Novig, ForecastEx, Robinhood, PredictIt (10% of profit), Predict.fun, Opinion, Limitless (blocks US), Myriad. ## Caution guide Fake Polymarket bot repos (StepSecurity, SlowMist reports) have stolen wallet keys. Use a burner wallet, read package.json and dependencies, run in an isolated VM, don't trust stars. Kalshi taker fee = 0.07 × p × (1 − p), max 1.75¢; Polymarket expanded taker fees on 30 March 2026, makers earn rebates; fetch /fee-rate. Venues may resolve the same event differently. Suggested path: watch free scanners, dry-run with Kalshi demo, backtest, go live small. Not financial advice. ## Independent Neural Bet Partner Guide (unofficial) This section is run by a Neural Bet partner and community. It is NOT the official Neural Bet website. Official site: https://www.neural-bet.com. Only official contact: @neuralbetsupport on Telegram (https://t.me/neuralbetsupport). DYOR. - Arbitrage engine for Polymarket and Kalshi: buys both sides when the combined price is under $1.00. - Official track record is updated daily on https://www.neural-bet.com. This site does not reproduce figures. - Official minimum deposit: $100, which activates profit sharing from arbitrage trading. - Deposits: USDC or USDT on Polygon (recommended, ~30 s), Ethereum or BNB Smart Chain, with 0% Neural Bet fees (network fees only). Tron, Solana, Base and Arbitrum are not credited. Bank card (Visa/Mastercard): 3% fee added, includes an Académie IA course, credited within minutes; no withdrawals for 30 days after a card payment. No maximum deposit. - Withdrawals: from $5, no maximum, no exit fee, in USDC on Polygon to your own wallet, usually within a few hours. The algorithm pauses and releases funds once in-progress arbitrages complete. - Eligibility: 18+ only; not available to US citizens, US residents or people subject to US sanctions. Users must accept the terms of use in Telegram. - Verification: every deposit and withdrawal can be checked on Polygonscan, BscScan or Etherscan; every arbitrage is listed in the app with gain and amount committed. - Guide chapters (https://predictionmarketarbitrage.com/neuralbet/guide): 01 Getting started, 02 Deposits, 03 Tracking, 04 Withdrawals, 05 Referral, 06 Your money, 07 Telegram in brief. - Funds: deposits pass through a shared deposit/withdrawal pool into operating capital. Only undeployed funds are held in a Safe multisignature vault; funds in arbitrage sit on the platforms. The founders (Lucas and Pierre) cannot individually move client funds. - Safeguards: max 1.5% of client funds per arbitrage; an unhedged leg is closed within 50 ms and new operations pause; a $20 loss on a pool stops the algorithm for all clients pending manual review (withdrawals stay available). A Morpho-style non-custodial setup is being rolled out, not live yet. Safeguards limit risk; they do not guarantee capital. - Community offers (a $250 USDC conditional bonus) are offered by our partner team in its own name, not by Neural Bet. - Sign up (partner link): https://t.me/Neuralbettingbot?start=ref455806787 - Not available to US citizens or residents. ## Pages https://predictionmarketarbitrage.com/ · /neuralbet · /neuralbet/guide · /neuralbet/etherfi · /neuralbet/videos · /glossary · /disclaimer The /neuralbet section (overview, guide, etherfi, videos) is the Independent Neural Bet Partner Guide and has its own section menu. ## Complete published article library Canonical site: https://predictionmarketarbitrage.com. Discovery index refreshed 2026-11-16. Neuralbet (Neural Bet) is covered by the independent PMA partner guide at https://predictionmarketarbitrage.com/neuralbet. PMA is not the official Neural Bet operator. These articles are educational, not trading recommendations or proof of attainable returns. ### Prediction Market Arbitrage: How a 3¢ Gap Shrinks After Fees Canonical: https://predictionmarketarbitrage.com/blog/3-cent-gap-after-fees Published: 2026-11-16 Updated: 2026-11-16 Author: PMA Research Summary: A 3-cent arbitrage gap between Kalshi and Polymarket looks profitable until taker fees eat the margin. Here is the math behind the break-even spread. A three-cent price difference between two prediction markets looks like free money on a screen. In reality, executing that trade requires crossing two order books, paying two sets of exchange fees, and taking on hidden execution risks. This guide breaks down the exact math of a cross-exchange arbitrage trade to show where the theoretical profit actually goes. ## Key takeaways * A gross arbitrage edge of 3 cents per contract often results in a net loss after applying standard taker fees on both platforms. * Kalshi charges a taker fee based on the expected earnings of a contract, peaking at roughly 1.75 cents per contract at the 50-cent mark for standard markets with a default multiplier. * Polymarket applies variable taker fee rates depending on the market topic, meaning traders must account for different rates across categories. * Using resting maker orders eliminates Polymarket fees but introduces leg risk, where one side of the trade fills and the other moves away. ## Why does a 3-cent prediction market arbitrage gap disappear? Arbitrage in prediction markets relies on buying opposing outcomes of the same event across different venues for a combined cost of less than $1.00. Since a winning YES contract pays $1.00 at resolution, securing both sides for 97 cents implies a guaranteed 3-cent gross profit. However, gross profit is an illusion. Exchanges charge fees to match orders. When you spot a fleeting price discrepancy, you must use market orders or marketable limit orders to capture it immediately. These are taker orders. Both Kalshi and Polymarket charge taker fees that scale with the probability of the contract. Because the fees are highest near the 50-cent mark, a mid-probability arbitrage trade faces the maximum possible fee burden. By the time both legs of the trade settle, the combined fees frequently exceed the initial 3-cent discount. Furthermore, venues can resolve the "same" event differently. A political market might rely on an official agency call on one exchange and a consensus of news networks on another. If the resolution criteria diverge, you could lose both sides of the trade. ## How do you calculate Kalshi and Polymarket taker fees? To understand the net outcome, you must calculate the exact fee for each leg. Both platforms use a parabolic fee structure that peaks at 50 cents and tapers off toward the extremes of 1 cent and 99 cents. The Kalshi taker fee formula is 0.07 multiplied by a contract multiplier (M), the number of contracts, the price, and one minus the price. As outlined in the [Kalshi: Fee Schedule PDF](https://kalshi.com/docs/kalshi-fee-schedule.pdf), the exchange rounds up such that the fee plus the position cost is rounded to a centicent, though printed fee tables often display to the cent. For a standard contract (M=1) priced at 50 cents, the unrounded math is 0.07 x 1 x 1 x 0.50 x 0.50, which equals 0.0175. Therefore, the maximum fee is about 1.75 cents per contract. Polymarket uses a similar formula but applies different multipliers based on the market category, while makers pay no fee and can earn rebates. According to the official [Polymarket: Fees](https://docs.polymarket.com/trading/fees) documentation, the crypto category carries a 0.07 rate, sports and economics carry a 0.05 rate, and politics carries a 0.04 rate. Geopolitics remains fee-free. The formula is the category rate multiplied by shares, price, and one minus the price. Polymarket rounds fees to five decimal places in USDC, not to the nearest cent. ## What is the net outcome of a 100-contract worked example? Consider a hypothetical scenario where you spot a pricing gap. You want to buy 100 YES contracts at 46 cents on Polymarket and 100 NO contracts at 51 cents on Kalshi. The gross cost is $46.00 on Polymarket and $51.00 on Kalshi, totaling $97.00. Since one side will inevitably pay out $100.00 at resolution, your gross edge is $3.00. Now apply the fees. On the Kalshi side, you are buying 100 contracts at 51 cents. The unrounded fee calculation is 0.07 x 1 x 100 x 0.51 x 0.49. This equals $1.7493, which the fee table displays as approximately $1.75. On the Polymarket side, assume this is an economics market with a 0.05 fee rate. The calculation is 0.05 x 100 x 0.46 x 0.54. This equals exactly $1.24200. Your total fees are $1.7493 plus $1.24200, equaling $2.9913 before venue rounding. You subtract this from your $3.00 gross edge. Your net profit is $0.0087 before venue rounding. You have tied up $99.9913 in capital (the $97.00 purchase price plus the $2.9913 in fees), taken on platform risk, and assumed resolution risk, all to earn less than a single penny. If the Polymarket leg was a crypto market with a 0.07 rate, the fee would be $1.73880, turning the trade into a net loss of $0.4881. ## What is the break-even gap formula for prediction markets? To avoid executing unprofitable trades, you need a reliable break-even formula. The required gap depends heavily on the price level and the specific Polymarket category rate. You can automate this math using a [fee calculator](/fee-calculator), but the manual formula requires adding the Kalshi fee and the Polymarket fee to the combined purchase price. Break-even occurs when: Price A + Price B + Fee A + Fee B = $1.00. The table below illustrates the hypothetical net outcome of a 100-contract trade with a gross $3.00 edge across different price levels, assuming a Polymarket rate of 0.05 and unrounded Kalshi fees for a standard M=1 contract. | Polymarket Price | Kalshi Price | Gross Edge | Polymarket Fee | Kalshi Fee | Net Outcome | | :--- | :--- | :--- | :--- | :--- | :--- | | $0.10 | $0.87 | $3.00 | $0.45000 | $0.7917 | +$1.7583 | | $0.30 | $0.67 | $3.00 | $1.05000 | $1.5477 | +$0.4023 | | $0.46 | $0.51 | $3.00 | $1.24200 | $1.7493 | +$0.0087 | | $0.70 | $0.27 | $3.00 | $1.05000 | $1.3797 | +$0.5703 | | $0.90 | $0.07 | $3.00 | $0.45000 | $0.4557 | +$2.0943 | Under the stated hypothetical prices and fees, the modeled margin is larger away from the midpoint. These figures are before Kalshi venue rounding and are not guaranteed executable returns. Slippage across depth further complicates this. If the top of the order book only has 20 contracts available at these prices, your remaining 80 contracts will fill at worse prices, instantly destroying the remaining margin. ## How does maker vs taker execution impact arbitrage risk? One way to bypass the heavy fee burden is to use resting limit orders. As noted in the [Polymarket Help Center: Trading Fees](https://help.polymarket.com/en/articles/13364478-trading-fees) guide, makers are never charged fees. If you place a maker order on Polymarket and wait for someone to trade against it, you save the entire taker fee. If you apply this to the previous 46-cent and 51-cent example, eliminating the $1.24200 Polymarket fee raises the modeled profit before Kalshi venue rounding from $0.0087 to $1.2507. However, this introduces severe leg risk. If you place a resting order on Polymarket and simultaneously take the 51-cent Kalshi order, the Polymarket price might move to 48 cents before your order fills. You are now holding an unhedged Kalshi position. If you try to automate this process, be extremely cautious regarding third-party software. Fake Polymarket arbitrage bot repos have stolen private keys from unsuspecting traders. Always review our [fake bot checklist](/fake-bot-checklist) before running unknown code. This is not financial advice, but protecting your capital is the first rule of arbitrage. ## Frequently asked questions ### Do geopolitical markets on Polymarket charge taker fees? No. Geopolitical and world events markets are currently fee-free on Polymarket. This makes arbitrage slightly more viable in this specific category, though Kalshi taker fees will still apply to the other leg of the trade. ### Can I avoid Kalshi fees by using maker orders? Kalshi does not have a universal blanket maker-free policy. While the default maker multiplier is 0 for many standard markets, the [Kalshi: Fee Schedule PDF](https://kalshi.com/docs/kalshi-fee-schedule.pdf) notes that there are many series exceptions. The general maker rate when M=1 is a quarter of the taker rate, and certain combo markets with M=2 charge maker fees at 50 percent of the standard taker rate. ### Why do cheap contracts yield higher net arbitrage profits? The fee formulas on both exchanges multiply the price by one minus the price. This creates a curve that peaks at 50 cents and drops significantly near 10 cents or 90 cents, leaving more of the gross edge intact at the extremes. ## Sources * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) * [Polymarket Help Center: Trading Fees](https://help.polymarket.com/en/articles/13364478-trading-fees) * [Kalshi: Fee Schedule](https://kalshi.com/fee-schedule) --- ### What Is Prediction Market Arbitrage? A Beginner's Primer Canonical: https://predictionmarketarbitrage.com/blog/what-is-prediction-market-arbitrage Published: 2026-11-15 Updated: 2026-11-16 Author: PMA Research Summary: Learn the YES/NO contract mechanics of prediction market arbitrage and the effects of fees, slippage, execution risk, and mismatched settlement rules. Prediction market arbitrage sounds like the ultimate free lunch: buying both sides of a binary event for less than their guaranteed payout. In reality, capturing these fleeting price gaps requires managing a minefield of execution risks, shifting fee structures, and mismatched resolution rules. This primer breaks down the mechanics of locking in a spread and why the spreadsheet math rarely survives contact with live order books. ## Key takeaways * Prediction market arbitrage involves buying complementary YES and NO contracts for a combined cost under $1.00. * Price discrepancies typically stem from fragmented liquidity, differing user bases, and varying regulatory environments across venues. * Gross spreads are frequently erased by taker fees, slippage, and the capital costs of locking up funds until resolution. * Beginners should start by observing live scanners and paper trading before risking capital on cross-venue execution. ## How does prediction market arbitrage actually work? To understand the strategy, you must first understand the underlying instruments. As noted by regulators, event contracts are typically structured as [swaps](https://www.cftc.gov/LearnandProtect/PredictionMarkets) that derive their value from the outcome of a specific real-world occurrence. These markets are often based on yes-no scenarios, allowing for only [two possible outcomes](https://www.cftc.gov/LearnandProtect/PredictionMarkets). Crucially, a YES contract usually pays exactly $1.00 at resolution if the event occurs, and $0.00 if it does not. The NO contract does the exact opposite. Because this framework usually guarantees a [fixed payout](https://www.cftc.gov/LearnandProtect/PredictionMarkets) of $1.00 for the winning side, holding both a YES and a NO contract for the exact same event guarantees a $1.00 return. Arbitrage occurs when you can acquire both sides for a combined cost of less than $1.00. Consider a hypothetical scenario where you spot a pricing inefficiency across two different platforms: | Contract Leg | Venue | Quoted Ask Price | Quantity | Total Cost | | :--- | :--- | :--- | :--- | :--- | | Buy YES | Platform A | $0.46 | 100 | $46.00 | | Buy NO | Platform B | $0.51 | 100 | $51.00 | | **Combined** | **Both** | **$0.97** | **100 pairs** | **$97.00** | In this hypothetical example, you spend $97.00 to acquire 100 complete pairs. At the [expiration](https://www.cftc.gov/LearnandProtect/PredictionMarkets) of the event, one side will inevitably win, paying out $100.00. The $3.00 difference represents your gross profit. ## Why do price gaps exist across different venues? If financial markets are highly efficient, you might wonder why these sub-dollar opportunities exist at all. The answer lies in market structure. First, liquidity is highly fragmented. Unlike traditional equities that trade on centralized consolidated tapes, prediction markets operate in silos. A surge of buying pressure on one platform does not automatically update the order book on another platform. Second, different venues attract different user bases with distinct biases and information sets. A crypto-native platform might heavily discount the probability of a specific regulatory action, while a traditional fiat-based exchange might price that same risk much higher. Finally, regulatory restrictions prevent seamless capital flow. Many institutional market makers cannot legally operate across both regulated US exchanges and offshore decentralized platforms. This friction prevents large players from instantly arbitraging away the price differences, leaving temporary windows open for retail traders. ## Why is prediction market arbitrage rarely free money? On a spreadsheet, buying a dollar for 97 cents is a flawless strategy. In live trading, four major frictions aggressively consume that gross margin. The most immediate hurdle is trading fees. Immediate execution generally requires marketable orders that take liquidity on each venue, but no order guarantees a cross-venue entry. Kalshi calculates its general taker fee as 0.07 x contracts x price x (1 - price), with a raw fee of 1.75 cents per contract at the 50-cent mark for a default-multiplier market, before the applicable venue rounding. Meanwhile, Polymarket charges taker fees on most categories, while makers pay no fee and can earn rebates. When you subtract these fees from a 3-cent gross spread, your net profit often drops to fractions of a penny. Execution risk, or "leg risk," is the second friction. You might successfully buy the YES contract at 46 cents, only to find the 51-cent NO contract has vanished before your second order routes. You are now holding an unhedged directional bet. Resolution risk is the most dangerous trap. Two venues can resolve the "same" event differently. One platform might rely on an Associated Press race call, while another requires official congressional certification. If the outcomes diverge, both of your contracts could resolve to $0.00. Finally, capital lockup degrades your annualized returns. Tying up $97.00 for six months to make a net profit of $0.50 is a terrible use of capital compared to the risk-free rate of standard treasury yields. ## What are the most common beginner mistakes? New traders consistently underestimate the impact of fees. They see a green indicator on a basic scanner, execute the trade, and realize their account balance shrank because the execution costs exceeded the gross spread. Always run your numbers through a reliable fee calculator before routing an order. Another frequent error is ignoring order book depth. A platform might quote a NO contract at 51 cents, but only have two contracts available at that price. If you try to buy 100 contracts, slippage will push your average fill price well above 51 cents, instantly destroying the arbitrage. Security lapses also plague newcomers. The allure of automated profits leads many to download unverified code. Fake Polymarket arbitrage bot repos have stolen private keys from affected users. Always review our fake bot checklist before granting API access or wallet permissions to any third-party software. ## What is a safe path for first steps? The most sustainable way to learn is through observation rather than immediate execution. Start by watching live market scanners without risking capital. Track how long spreads stay open and watch how order book depth fluctuates during breaking news events. Next, move to paper trading. Record your hypothetical entries, being brutally honest about the prices you could actually hit as a taker. Calculate the exact fees for both platforms and factor in the time your capital would be locked up. If you decide to transition to live markets, use a demo environment if the exchange provides one. When you finally trade with real capital, use the smallest possible size. The goal of your first 50 trades is not to generate meaningful income, but to prove that your execution mechanics work and that your fee calculations are accurate. *Disclaimer: This guide is for educational purposes only and does not constitute financial advice.* ## Frequently asked questions ### Can I use limit orders to avoid taker fees? Yes, using limit orders makes you a maker if the order is not immediately marketable, which often incurs zero fees or even earns rebates. However, resting a limit order means you are not guaranteed a fill, which introduces massive leg risk if only one side of your arbitrage executes. ### How fast do I need to be to catch these spreads? Cross-venue spreads caused by breaking news often close in milliseconds as automated bots sweep the books. Slower, structural spreads caused by differing user biases can stay open for minutes or hours, but they usually offer much thinner margins. ### Do I have to hold the contracts until the event resolves? No. If the market prices converge before the event concludes, you can sell both legs early to unlock your capital. However, exiting early means paying a second round of trading fees, which must be factored into your initial profit calculations. ## Sources * [CFTC: Understanding Prediction Markets and Event Contracts](https://www.cftc.gov/LearnandProtect/PredictionMarkets) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) --- ### Measuring Prediction Market Accuracy: Brier Scores Canonical: https://predictionmarketarbitrage.com/blog/measuring-prediction-market-calibration Published: 2026-11-13 Updated: 2026-11-16 Author: PMA Research Summary: Learn how to measure prediction market accuracy using Brier scores and calibration curves to find systematic pricing differences across trading venues. Prediction markets promise to distill crowd wisdom into precise probabilities, but taking those prices at face value without measuring their historical accuracy is a recipe for negative expected value. By applying rigorous scoring rules like the Brier score and plotting calibration curves, quantitative traders can identify exactly where venues systematically misprice risk. Understanding these metrics is the first step to finding structural differences between platforms. ## Key takeaways * The Brier score is the standard proper scoring rule for binary forecasts, where a score of 0.0 indicates perfect accuracy. A constant 0.5 forecast always scores 0.25, though the expected score for a base-rate constant forecast is q*(1-q). * Reliability diagrams visually expose systematic biases, such as the tendency for long-shot contracts priced below 10 cents to be structurally overvalued. * Academic working papers suggest that prediction market calibration generally improves as trading volume and unique trader counts increase, though this is an association rather than a universal monotonic effect. * Differences in calibration between venues can signal cross-exchange opportunities, provided traders account for distinct fee structures, rounding precision, and resolution rules. ## What is a Brier score and how do you calculate it? The Brier score measures the mean squared difference between predicted probabilities and actual binary outcomes. For a YES contract that pays $1.00 at resolution, the outcome is 1 if the event occurs and 0 if it does not. A lower score is always better. If a market prices an event at 50 cents (0.50 probability) and you always guess 0.50 against a 50/50 base rate, your Brier score is 0.25. Consider a worked example with three resolved markets: 1. Market A: Priced at 0.80, resolved YES (1). 2. Market B: Priced at 0.30, resolved NO (0). 3. Market C: Priced at 0.90, resolved NO (0). The squared errors are: * Market A: (0.80 - 1.0)^2 = 0.04 * Market B: (0.30 - 0.0)^2 = 0.09 * Market C: (0.90 - 0.0)^2 = 0.81 The average of these squared errors is (0.04 + 0.09 + 0.81) / 3 = 0.313333. This score is heavily penalized by Market C's confident but incorrect forecast. Here is a simple Python snippet using the standard library to calculate this metric: ```python def calculate_brier_score(predictions, outcomes): if len(predictions) != len(outcomes): raise ValueError("Lists must be the same length") squared_errors = [(p - o) ** 2 for p, o in zip(predictions, outcomes)] return sum(squared_errors) / len(predictions) preds = [0.80, 0.30, 0.90] results = [1, 0, 0] print(f"Brier Score: {calculate_brier_score(preds, results):.6f}") ``` ## How do reliability diagrams reveal mispriced probabilities? While a single score summarizes overall accuracy, calibration curves (or reliability diagrams) reveal the shape of a market's forecasting ability. To build one, analysts group market prices into bins (for example, 0.0 to 0.10, 0.10 to 0.20) and plot the average predicted probability on the x-axis against the actual observed frequency of YES resolutions on the y-axis. Perfect calibration results in a straight 45-degree line. If the curve dips below the diagonal at the low end, it means long-shot contracts are overpriced. For instance, if contracts priced at 5 cents only resolve YES 1 percent of the time, buying them is a losing proposition. As noted in [scikit-learn documentation](https://scikit-learn.org/stable/modules/model_evaluation.html#brier-score-loss), calibration improves the Brier score and log loss by making predicted probabilities more accurate for allocation decisions under uncertainty. ## Are Kalshi and Polymarket prices actually well calibrated? A [working paper on calibration](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7355520) studies Kalshi forecasts across horizons. Its findings should be interpreted within the sample, weighting, and forecast times actually studied, not as a universal venue score. Compare models on the same events and horizons; a late forecast made after most uncertainty has resolved is not directly comparable with an early forecast. Similarly, domain-specific dynamics play a massive role. A preprint on [domain-specific calibration dynamics](https://arxiv.org/html/2602.19520) highlights that political markets often see prices systematically compressed toward 50 percent, while short-horizon weather markets can be mildly over-extreme. Entertainment and finance markets tend to track the perfect-calibration diagonal much more closely. This means a trader relying on a [fake bot checklist](/fake-bot-checklist) to filter out noise must also adjust their baseline expectations depending on the specific category they are trading. ## How does trading volume impact prediction market accuracy? Liquidity and participation are the lifeblood of market efficiency. The data shows a general association between the depth of a market and its forecasting accuracy, though this is not a universal monotonic rule. According to [Kalshi's internal research](https://kalshi.com/research/kalshi-research-calibration.pdf), calibration generally improves alongside volume across time horizons. Trader count is also associated with accuracy, though depth of participation only improves calibration up to a limit set by genuine uncertainty; no amount of trading fully substitutes for the passage of time. However, depth of participation only improves calibration up to a limit set by genuine uncertainty; no amount of trading fully substitutes for the passage of time. ## Where do calibration differences create opportunities? When two venues exhibit different calibration curves for the same underlying event, quantitative traders can construct market-neutral positions. If a political market on Polymarket is structurally compressed toward 50 cents while the equivalent market on Kalshi tracks the 45-degree line perfectly, a trader might buy the underpriced favorite on Kalshi and hedge by buying the underpriced underdog on Polymarket. However, executing this requires strict attention to trading costs and rules. You must run the numbers through a [fee calculator](/fee-calculator) before deploying capital. Kalshi charges a general taker fee calculated as `round up(M * 0.07 * C * P * (1 - P))`, where the default multiplier M=1. While printed fee tables round to cents, the July 7, 2026 effective schedule states that the fee plus position cost is rounded up to a centicent. Traders must explicitly read current precision from live fee parameters and should not assume a per-contract 1.75-cent hard cap across multipliers. Furthermore, Kalshi's general maker rate when M=1 is a quarter of the taker rate, not half, and there is no universal blanket maker-free claim. Meanwhile, Polymarket expanded taker fees to nearly every category on March 30, 2026. Their fee formula is `fee = C * feeRate * p * (1 - p)`. Rates vary by category: Crypto is 0.07; sports, economics, culture, weather, and other are 0.05; finance, politics, mentions, and tech are 0.04; geopolitics is 0; and combos are 0.05. Makers are never charged, though rebate eligibility does not guarantee revenue. Crucially, Polymarket fees are rounded to five decimal places in USDC, not cents. For example, 100 shares at a 0.46 price with a 0.05 rate incurs a fee of 1.24200 USDC, not exactly 1.24. Furthermore, venues can resolve the "same" event differently based on their specific rulebooks. A discrepancy in calibration might not be an arbitrage opportunity at all, but rather a rational pricing of different settlement criteria. Always read the fine print. This is not financial advice. ## Frequently asked questions ### What is the difference between Brier score and log loss? The Brier score measures the mean squared error of probability forecasts, making it highly sensitive to large deviations. Log loss (cross-entropy loss) heavily penalizes confident but wrong predictions by using the natural logarithm, meaning a prediction of 0.99 that resolves NO will result in a massive penalty compared to the Brier score. ### Why do long-shot bias anomalies exist in prediction markets? Long-shot bias occurs because traders often overvalue low-probability events, treating cheap contracts like lottery tickets. This structural inefficiency causes contracts priced below 10 cents to resolve YES far less frequently than their implied probability suggests, creating a drag on overall market calibration. ### Can I compare calibration directly between Polymarket and Kalshi? Direct comparison is difficult because the platforms list different types of markets and attract different user bases. However, by filtering for matched markets with similar time horizons and trading volumes, analysts can isolate venue-specific pricing behaviors and identify structural differences in how their respective crowds assess risk. ## Sources * [Scikit-Learn: Brier Score Loss](https://scikit-learn.org/stable/modules/model_evaluation.html#brier-score-loss) * [SSRN: Calibration in Prediction Markets: Theory and Evidence by Nicole Kagan, Rubens Baiocchi (Working Paper)](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7355520) * [Arxiv: Decomposing Crowd Wisdom: Domain-Specific Calibration Dynamics in Prediction Markets (Preprint)](https://arxiv.org/html/2602.19520) * [Kalshi: Calibration in Prediction Markets: Theory and Evidence](https://kalshi.com/research/kalshi-research-calibration.pdf) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) --- ### From Intrade to Kalshi: Prediction Market Arbitrage History Canonical: https://predictionmarketarbitrage.com/blog/evolution-of-cross-venue-arbitrage Published: 2026-11-13 Updated: 2026-11-16 Author: PMA Research Summary: Explore the evolution of prediction market arbitrage, from the Iowa Electronic Markets and Intrade to modern API-driven trading on Kalshi and Polymarket. Prediction market arbitrage was once a manual, slow-moving pursuit where traders clicked through clunky interfaces to lock in pennies. Today, it is a high-frequency domain dominated by algorithmic execution across regulated exchanges and decentralized protocols. Understanding how we transitioned from academic experiments to institutional-grade order books is essential for anyone looking to extract edge from event contracts. ## Key takeaways * The Iowa Electronic Markets proved the concept of event contracts in 1988, but strict regulatory caps prevented early arbitrageurs from scaling their strategies. * Offshore platforms like Intrade offered massive liquidity in the 2000s before aggressive Commodity Futures Trading Commission (CFTC) enforcement forced their closure. * Modern venues like Kalshi and Polymarket have institutionalized the asset class, replacing manual clicking with API-driven market making. * Arbitrage spreads have compressed significantly as tooling improved, forcing traders to account for complex fee structures, API rate limits, and venue-specific resolution rules. ## How did early prediction markets like IEM and Intrade operate? The modern era of event contracts began in 1988 with the Iowa Electronic Markets (IEM). Operated by the University of Iowa, the IEM was designed to test whether a market of real-money traders could forecast presidential elections more accurately than traditional polling. The CFTC granted the IEM a no-action letter in 1992, allowing it to open to the public under strict conditions, including a $500 investment cap and a non-profit research mandate. While the IEM successfully demonstrated the predictive power of financial incentives, its low limits made it impossible for professional traders to deploy meaningful capital or execute cross-venue arbitrage. As the internet matured, offshore platforms emerged to fill the demand for higher-stakes trading. The most prominent was Intrade, a Dublin-based exchange founded in 2001. Alongside the UK-based sports betting giant Betfair, Intrade popularized political and current-event contracts for a global audience. Intrade operated without the strict position limits of the IEM, allowing traders to move tens of millions of dollars during the 2008 and 2012 US presidential elections. However, this unregulated growth attracted regulatory scrutiny. In 2012, the CFTC sued Intrade for offering unregistered options to US residents. The legal pressure forced Intrade to block American users, and the platform ultimately shut down in 2013 following a separate solvency crisis. ## What was the impact of the PredictIt no-action letter? Following the collapse of Intrade, the US market was left with a massive liquidity vacuum. In 2014, a new platform called PredictIt launched under a CFTC no-action letter. Operated as a partnership between Victoria University of Wellington and a US political consulting firm, PredictIt was modeled on the IEM's academic exemption but allowed for slightly larger stakes, historically capping investments at $850 per contract per market before raising the limit to $3,500 on July 15, 2025. For years, PredictIt was the primary venue for US political bettors, but its high fees and strict position limits severely constrained arbitrage opportunities. Traders often found themselves unable to hedge positions effectively due to the historical $850 cap. Furthermore, PredictIt charges a 10% fee on trading profits. While the platform removed its 5% withdrawal fee on September 23, 2026, and offers a promotional waiver on the next credit card deposit up to $1,000 (ACH deposits are free, subsequent card deposits incur a 3% fee), the profit fee remains a significant hurdle for high-frequency arbitrage. The regulatory peace was also short-lived. In August 2022, the CFTC revoked PredictIt's no-action letter, alleging the platform had violated the terms of its exemption. PredictIt challenged the revocation in court, leading to a protracted legal dispute that allowed the platform to continue operating through the 2024 election cycle with a limited contract offering. ## How did Polymarket and Kalshi change the regulatory environment? The 2020s marked a structural shift in how prediction markets operate, driven by two distinct approaches to regulation and technology. Polymarket launched in 2020 as a crypto-native platform built on the Polygon network. By settling trades in stablecoins, Polymarket eliminated the volatility risk associated with earlier blockchain experiments. However, its borderless approach quickly drew the ire of US regulators. In January 2022, the CFTC fined Polymarket $1.4 million for operating an unregistered futures exchange and ordered it to block US customers. Conversely, Kalshi took the traditional regulatory route. In an order dated November 3, 2020 (announced November 4), the CFTC approved Kalshi as a Designated Contract Market (DCM), making it the first federally regulated exchange built specifically for event contracts. Kalshi initially focused on economic data and weather, but its ambitions were much larger. After the CFTC blocked Kalshi from listing contracts on congressional control in 2023, the exchange sued the regulator. In September 2024, a federal court ruled in Kalshi's favor, paving the way for the platform to list US election contracts just weeks before the vote. This ruling fundamentally altered the industry, bringing massive institutional liquidity to onshore markets. ## Timeline of prediction market evolution | Year | Event | Impact on Arbitrage and Trading | | :--- | :--- | :--- | | 1988 | Iowa Electronic Markets launches | Proves the concept of event contracts; low caps prevent scaling. | | 2001 | Intrade founded in Dublin | Introduces high-liquidity political trading; creates early cross-border spreads. | | 2012 | CFTC sues Intrade | US enforcement restricts access; the subsequent shutdown should not be reduced to a single proven cause. | | 2014 | PredictIt receives no-action letter | Revives US political trading but limits arbitrage with an $850 cap. | | 2020 | Kalshi approved as a DCM | Adds a CFTC-designated exchange focused on event contracts. | | 2020 | Polymarket launches on Polygon | Introduces low-fee, stablecoin-settled decentralized trading. | | 2022 | CFTC fines Polymarket $1.4M | Forces Polymarket to geofence US users; offshore liquidity pools grow. | | 2024 | Federal court rules for Kalshi | A key court ruling concerning Kalshi election contracts; not blanket permission for every product or user. | ## How have APIs and fees compressed arbitrage spreads? The transition from retail-dominated platforms like PredictIt to API-first exchanges like Kalshi and Polymarket has fundamentally changed how traders extract value. In the past, a trader might manually buy a YES contract on one site and sell it on another, pocketing a 5-cent spread. Today, automated bots scan order books millisecond by millisecond, compressing spreads to fractions of a cent. If you are building automated systems, reviewing a [fake bot checklist](/fake-bot-checklist) is critical to avoid deploying flawed logic. To execute profitable cross-venue arbitrage today, quants must rigorously model fee structures. Consider a hypothetical scenario where a YES contract pays $1.00 at resolution. You spot a YES contract trading at 48 cents on Kalshi and 52 cents on Polymarket. On Kalshi, the general taker fee formula is `round up(M * 0.07 * C * P * (1 - P))`, where M is the multiplier (default taker M=1), C is the number of contracts, and P is the price. This fee, plus the position cost, is rounded up to a centicent. For 100 contracts at 50 cents, the unrounded fee is 1.75 cents per contract, but traders must explicitly read current precision from live fee parameters rather than assuming a hard cap. Kalshi also charges maker fees on many series, typically using the formula `round up(M * 0.0175 * C * P * (1 - P))`, meaning there is no universal blanket maker-free claim. Meanwhile, Polymarket charges taker fees based on the category, using the formula `fee = C * feeRate * p * (1 - p)`. Crypto markets charge a 0.07 rate, sports/economics/culture/weather/other charge 0.05, finance/politics/mentions/tech charge 0.04, and geopolitics are fee-free. Combos charge 0.05. Polymarket fees are rounded to exactly five decimal places in USDC, not cents. For example, 100 shares at a 0.46 price with a 0.05 rate results in a fee of 1.24200 USDC. Makers on Polymarket are never charged fees and may be eligible for rebates, though this is not guaranteed revenue. If you cross the spread as a taker on both venues, the combined fees will severely eat into your gross margin. Using a precise [fee calculator](/fee-calculator) is mandatory before deploying capital. Furthermore, traders must account for resolution risk. Venues can resolve the "same" event differently based on their specific rulebooks. Polymarket uses the UMA optimistic oracle for prediction markets (which includes a 2-hour challenge period where disputes can take days) and Chainlink TWAP for up-down markets. Kalshi relies on its own specified source agencies. A contract on "US GDP Growth" might settle based on the advance estimate on one exchange and the final estimate on another. Cross-venue opposite labels alone do not guarantee a payout. If you ignore these nuances, what looks like a theoretical arbitrage is actually an unhedged directional bet. This is not financial advice, but failing to read the fine print is the fastest way to lose money in event contracts. ## Frequently asked questions ### What is prediction market arbitrage? Prediction market arbitrage involves buying a contract on one exchange and simultaneously selling the opposing contract on another exchange to attempt to lock in a theoretical profit. This strategy exploits price discrepancies between different platforms offering markets on the exact same event, though execution and resolution risks mean there is no guaranteed profitability. ### Why do spreads exist between platforms? Spreads exist due to fragmented liquidity, differing user demographics, and varying regulatory restrictions. For example, a platform restricted to non-US users might price a US political event differently than a fully regulated US exchange due to different information flows and capital constraints. ### Are prediction markets legal in the US? Yes, but they are strictly regulated by the Commodity Futures Trading Commission (CFTC). Platforms like Kalshi operate legally as Designated Contract Markets, while others operate under specific exemptions or face restrictions on serving US customers. It is important to note that CFTC regulation is not FDIC insurance, and trading event contracts does not provide risk-free capital. ## Sources * [National Bureau of Economic Research (Working Paper): ARE PREDICTION MARKETS POLITICALLY BIASED?](https://www.nber.org/system/files/working_papers/w35846/w35846.pdf) * [National Bureau of Economic Research (Working Paper): LIMITS TO ARBITRAGE AND PREDICTION MARKET EFFICIENCY: THE CASE OF PREDICTIT](https://www.nber.org/system/files/working_papers/w35845/w35845.pdf) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) * [PredictIt: deposits and withdrawals update](https://www.predictit.org/stories/detail/102/Product-Update--Deposits-and-Withdrawals-are-Free) * [PredictIt: platform announcements, July 15, 2025](https://www.predictit.org/platform-announcements) * [CFTC: Kalshi designation](https://www.cftc.gov/PressRoom/PressReleases/8302-20) --- ### 7 Recurring Mispricing Patterns in Prediction Markets Canonical: https://predictionmarketarbitrage.com/blog/recurring-prediction-market-mispricing-patterns Published: 2026-11-12 Updated: 2026-11-16 Author: PMA Research Summary: Discover seven recurring mispricing patterns in prediction markets, from longshot bias to capital lockup discounts, and how quants analyze them. Prediction markets are frequently marketed as perfectly efficient forecasting engines, yet their live order books reveal persistent structural anomalies. Capital constraints, retail behavioral biases, and fragmented liquidity create recurring mispricing patterns that quantitative traders can systematically identify. Exploiting these gaps requires distinguishing genuine arbitrage from disguised directional risk. ## Key takeaways * Multi-outcome markets frequently sum well above 100 percent due to retail traders' aversion to shorting and persistent overpricing of low-probability longshots. * Long-dated contracts on highly probable events trade at a discount because collateral is locked without yield, forcing prices down to reflect the risk-free rate. * Cross-venue price differences rarely represent true arbitrage, as subtle variations in resolution criteria or settlement rules usually explain the spread. * Automated trading in thin overnight markets carries severe risks, and traders must be wary of malicious software when seeking execution advantages. ## Why do multi-outcome prediction markets sum above 100 percent? In a perfectly efficient multi-outcome market where events are mutually exclusive and exhaustive, the implied probabilities of all outcomes must sum to exactly 100 percent. Executable YES asks can sum above 100% because of spreads and other frictions; the size of any excess is sample-specific and does not by itself establish a tradable arbitrage. This structural overpricing is driven by two distinct behavioral phenomena: short aversion and the favorite-longshot bias. Retail participants generally prefer buying a YES contract to shorting an outcome via a NO contract. This directional preference creates upward price pressure across the board. Academic working papers highlight this inefficiency; a preprint study on PredictIt found that bid prices on multi-outcome markets persistently summed to between 110 and 125, creating theoretical arbitrage opportunities even after fees [NBER: Limits to Arbitrage and Prediction Market Efficiency: The Case of PredictIt](https://www.nber.org/system/files/working_papers/w35845/w35845.pdf). Note that working papers linked on NBER are not necessarily peer-reviewed. Compounding this is the persistent overpricing of longshots. Traders overvalue low-probability events, treating cheap contracts like lottery tickets. Recent preprint analysis of Polymarket order books explores this favorite-longshot bias, suggesting that purchases below 10 cents lose 19.3 cents per dollar on average, while purchases at or above 90 cents yield a slight positive return [arXiv: The Favorite-Longshot Bias in Prediction Markets: Evidence from Polymarket](https://arxiv.org/html/2609.12878v1). When a market features a dozen unlikely candidates, the aggregate overpricing of these tail events pushes the total market sum well above 100 percent. ## How does capital lockup discount long-dated markets? When you purchase a prediction market contract, your capital is typically locked until the market resolves. In traditional finance, idle capital earns the risk-free rate. In prediction markets, fully collateralized positions usually earn zero yield. This opportunity cost forces long-dated contracts to trade at a discount to their true probability. Consider a worked numeric example for a contract resolving in exactly one year: * True probability of the event: 95 percent. * Risk-free rate (e.g., 1-year Treasury yield): 5.0 percent. * Undiscounted expected payout under the assumed probability: $0.95. * Opportunity cost discount: $0.95 / 1.05 = $0.9047. If you buy this YES contract at 90.47 cents, you are not capturing 4.53 cents of predictive edge. You are simply earning the risk-free rate on your locked collateral. A YES contract pays $1.00 at resolution, meaning your 90.47 cent investment yields roughly 5 percent over the year. Traders who ignore this cost of capital frequently mistake fair-value financing discounts for mispriced favorites. ## What causes weekend and overnight liquidity gaps? Prediction markets operate continuously, but liquidity provision does not. Market makers frequently widen their spreads or pull resting orders during weekends and overnight hours to protect against adverse selection when their monitoring systems are less active. This creates thin order books where small market orders can cause massive price slippage. This liquidity gap is particularly dangerous in post-resolution but unsettled markets. When an event occurs on a Saturday, the outcome may be known to the public, but the official oracle or settlement authority might not process the resolution until Monday morning. During this window, the market is technically still trading. Opportunistic traders might attempt to buy the winning YES contracts at 98 or 99 cents from stale resting orders. However, the capital required to sweep these pennies is substantial, and the risk of a delayed or disputed settlement remains. It is important to remember that a limit order can be a taker if marketable; only resting or non-marketable orders provide liquidity. ## Why do duplicate markets across venues price differently? Seeing the "same" event trading at 40 cents on one exchange and 50 cents on another looks like free money. It almost never is. Venues can resolve the "same" event differently based on their specific rulebooks. One platform might require an official government press release, while another relies on a consensus of three major news networks. One might use a UTC deadline, while the other uses Eastern Standard Time. If you attempt to arbitrage this spread by buying the cheap YES and selling the expensive YES, you are taking basis risk, not executing an arbitrage. You could easily lose both legs of the trade if the event occurs in a way that triggers one platform's YES criteria but fails the other's. Cross-venue opposite labels alone do not guarantee payout. Traders attempting to automate the scanning of these cross-venue discrepancies must exercise extreme caution regarding their infrastructure. The ecosystem is rife with malicious actors distributing compromised code. Numerous fake Polymarket arbitrage bot repos have stolen private keys from unsuspecting users who ran unverified scripts. Always consult a [fake bot checklist](/fake-bot-checklist) before deploying third-party trading software. ## How do bracket and range markets become inconsistent? Bracket markets, where an underlying metric like inflation or a candidate's vote share is divided into mutually exclusive ranges, frequently exhibit internal logical inconsistencies. A market pricing "Inflation above 3.0 percent" might trade at 40 cents, while the sum of the individual brackets "3.0 to 3.2 percent", "3.2 to 3.4 percent", and "Above 3.4 percent" might trade at 35 cents. These inconsistencies arise because liquidity is fragmented across multiple order books, and market makers may not perfectly synchronize their quotes across all related derivatives. However, capturing this spread requires executing multiple legs simultaneously. If you lift the offer on one bracket, the offers on the remaining brackets may be pulled before your subsequent orders route. Furthermore, transaction costs can quickly erode the theoretical margin. You must rigorously model your execution costs using a [fee calculator](/fee-calculator). For example, the Kalshi general taker fee formula is round up(M * 0.07 * C * P * (1 - P)), where the default taker multiplier M is 1. Explicitly, the current precision must be read from live fee parameters; traders should not assume a per-contract 1.75-cent hard cap after rounding or across multipliers. A general fee for 100 contracts at 0.51 is an unrounded 1.7493, though tables may display about 1.75. Meanwhile, Polymarket charges taker fees across most categories, calculated as fee = C * feeRate * p * (1 - p). The rates vary: Crypto is 0.07, sports, economics, culture, weather, and other are 0.05, finance, politics, mentions, and tech are 0.04, geopolitics is 0, and combos are 0.05. These fees are rounded to five decimal places in USDC, not cents. For instance, 100 shares at a 0.46 price with a 0.05 rate equals exactly 1.24200 USDC, not 1.24. Makers are never charged fees, though rebate eligibility does not guarantee revenue. If your bracket arbitrage yields 3 cents of gross edge but costs 4 cents in taker fees to leg into, the trade will result in a loss. Rigorous fee modeling is mandatory for any structural trading strategy. ## How should quants validate apparent pricing patterns? Statistical mispricing requires consistent sampling and genuinely out-of-sample evaluation. An average historical pattern is not a payoff guarantee. When analyzing historical market data to find mispriced contracts, quants look at calibration—how closely the market's implied probabilities match the actual frequency of outcomes. However, finite-sample calibration requires confidence intervals, as the Brier score includes components of resolution and uncertainty as well as reliability. It is crucial to understand that historical calibration bias is not riskless arbitrage. Just because a market has historically mispriced events in a certain probability bucket does not mean a trader can extract guaranteed profits, as future events may not follow the same distribution and transaction costs must be paid. ## Frequently asked questions ### What is the favorite-longshot bias? It is a persistent pricing anomaly where low-probability events trade at prices higher than their actual likelihood of occurring, while high-probability events trade at a discount. Retail traders overpay for longshots because they view them as cheap lottery tickets with high potential payouts. ### Why do long-dated prediction markets look cheap? Long-dated markets require traders to lock up their collateral for extended periods without earning interest. The apparent discount on highly probable events is usually just the market pricing in the risk-free rate of return that the capital could have earned elsewhere. ### Is cross-venue arbitrage risk-free? No. Cross-venue arbitrage carries significant basis risk because different platforms often have subtly different resolution criteria, oracle mechanisms, or settlement deadlines. What appears to be a price discrepancy is usually the market correctly pricing the difference in the underlying contract rules. ## Sources * [NBER: Limits to Arbitrage and Prediction Market Efficiency: The Case of PredictIt](https://www.nber.org/system/files/working_papers/w35845/w35845.pdf) * [arXiv: The Favorite-Longshot Bias in Prediction Markets: Evidence from Polymarket](https://arxiv.org/html/2609.12878v1) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) --- ### Partisan Bias and Herding: Behavioral Edges Between Venues Canonical: https://predictionmarketarbitrage.com/blog/partisan-bias-herding-prediction-markets Published: 2026-11-11 Updated: 2026-11-16 Author: PMA Research Summary: Explore how partisan wishful thinking, herding, and whale-driven moves create persistent cross-venue gaps in prediction markets like Polymarket and Kalshi. Prediction markets are often hailed as objective truth engines, but beneath the surface, they are driven by human traders with distinct biases, bankrolls, and political leanings. When you compare a crypto-native, international platform like Polymarket to a US-regulated venue like Kalshi or the political die-hards on PredictIt, you are not just comparing interfaces—you are comparing entirely different psychological ecosystems. Understanding these behavioral edges is crucial for arbitrageurs who want to exploit cross-venue gaps without getting trapped by the assumption that prices will inevitably converge. ## Key takeaways * **Demographics drive divergence:** Different platforms attract different user bases; Polymarket's crypto-native whales and PredictIt's capped retail traders create distinct pricing environments. * **Wishful thinking is real but nuanced:** While broad studies show no systematic left-right bias over long horizons, short-term partisan sentiment and "wishful thinking" can temporarily distort prices, especially in close races. * **Whales and herding create momentum:** Large, concentrated bets (like the 2024 "French whale") can trigger herding behavior, pushing prices away from fundamentals and creating potential, though not guaranteed, arbitrage opportunities. * **Structural limits prevent correction:** Position caps on platforms like PredictIt limit the ability of rational arbitrageurs to correct mispricing, allowing behavioral biases to persist longer than they would in uncapped markets. ## How Demographics Shape the Tape Prediction markets are not monolithic. The traders on Polymarket, who fund their accounts with USDC and often trade in massive size, have a different risk profile and demographic makeup than the retail traders on Kalshi or PredictIt. Research has consistently shown that prediction market traders are not representative of the general voting population. They tend to skew younger, more male, and more highly educated. However, a comprehensive [NBER working paper](https://www.nber.org/papers/w35846) analyzing 145 years of election betting data found no systematic overpricing of either left- or right-leaning candidates across the broad sweep of history. Yet, when we zoom in on specific platforms and shorter timeframes, behavioral quirks emerge. For instance, a study on the [2020 US Presidential Election](http://www.efmaefm.org/0EFMAMEETINGS/EFMA%20ANNUAL%20MEETINGS/2025-Greece/papers/Predictit3.pdf) found a direct link between partisan sentiment expressed on PredictIt message boards and asset mispricing, particularly for Republican assets in close races. This suggests that while the *market* might not be biased long-term, *pockets of traders* certainly are, and their "wishful thinking" can move prices. ## The Favorite-Longshot Bias and Partisan Sentiment One of the most persistent behavioral anomalies in prediction markets is the favorite-longshot bias, where traders overvalue unlikely outcomes and undervalue highly probable ones. This has been documented extensively, including in a recent analysis of [Kalshi's prediction market](https://cepr.org/voxeu/columns/economics-kalshi-prediction-market), which found that cheap contracts win less often than their price implies, while expensive contracts win slightly more often. This bias can be exacerbated by partisan sentiment. On platforms with position limits, like PredictIt (which historically capped positions at $850 per contract before raising it to $3,500 in July 2025), rational traders cannot deploy enough capital to correct the overpricing of longshots driven by enthusiastic partisans. As noted in another [NBER working paper on PredictIt](https://www.nber.org/system/files/working_papers/w35845/w35845.pdf) studying the 2016-2022 period, this structural limit led to significant pricing biases, with the sum of bid prices in multi-outcome markets often exceeding 110 cents on the dollar. ### A Worked Example: The Arbitrage Trap Imagine a hypothetical scenario where a YES contract pays $1.00 at resolution. * **Polymarket (Uncapped):** Candidate A is trading at 55 cents. * **PredictIt (Capped):** Candidate A is trading at 62 cents, driven by retail partisan enthusiasm. An arbitrageur might buy on Polymarket and sell on PredictIt, expecting the 7-cent gap to close. However, because PredictIt limits position sizes, the arbitrageur cannot short enough volume to force the price down. If evaluating a hypothetical current-fee scenario, one must also account for Polymarket's category-specific taker fees (expanded March 30, 2026) and Kalshi's general taker fee (M x 0.07 x contracts x price x (1 - price) before venue rounding). If the gap doesn't close before the event, the arbitrageur is forced to hold to resolution. This introduces a critical risk: venues can resolve the "same" event differently based on their specific rulebooks. If you are relying on a [fee calculator](/fee-calculator) to project profits, you must account for the possibility that the gap persists and the resolution criteria diverge. ## Whales, Herding, and Momentum In uncapped markets, the dynamics shift from retail wishful thinking to whale-driven momentum. The most famous recent example is the "French whale" on Polymarket, who reportedly earned $80 million betting on Donald Trump's 2024 victory. When a whale takes a massive position, it doesn't just move the price; it changes the narrative. Retail traders, assuming the whale possesses superior information, often herd into the same trade. This can push prices far beyond what polling or fundamental analysis would suggest. However, research on the [Iowa Electronic Markets](https://afajof.org/management/viewp.php?n=79184) (IEM) shows that while many traders are prone to biases like the disposition effect (holding losers too long) and the endowment effect, markets can still achieve efficient pricing. This efficiency arises because a small number of highly rational, price-determining traders exploit the mistakes of the biased majority. ## The Limits of Arbitrage So, how do these behavioral edges create persistent cross-venue gaps? It comes down to the limits of arbitrage. 1. **Capital Constraints:** Position limits on regulated platforms prevent arbitrageurs from fully correcting mispricing. 2. **Friction:** Fees, withdrawal costs, and the opportunity cost of locking up capital deter traders from exploiting small gaps. PredictIt removed its 5% withdrawal fee in September 2026 (ACH is free, subsequent card deposits incur a 3% fee with one waiver up to $1,000), but retains a 10% fee on trading profits. 3. **Resolution Risk:** The fear that platforms will interpret the outcome differently keeps spreads wider than they theoretically should be. When trading across venues, you are not just betting on the outcome of an event; you are betting on the behavior of the specific crowd on that platform. A crypto-native trader on Polymarket might anchor on a round number or follow a whale, while a retail trader on PredictIt might trade purely on partisan emotion. *(Note: This is an analysis of market mechanics, not financial advice.)* ## Navigating the Behavioral Landscape To succeed in prediction market arbitrage, you must look beyond the raw probabilities and understand the psychological forces driving the tape. * **Identify the dominant cohort:** Is the market driven by crypto whales, political partisans, or casual retail traders? * **Account for structural limits:** Are position caps preventing rational price discovery? * **Watch for herding:** Is a sudden price move driven by new information, or just retail traders following a large bet? By recognizing these behavioral edges, you can better assess whether a cross-venue gap is a genuine arbitrage opportunity or a trap set by the unique psychology of the platform's user base. For more insights on navigating these markets, check out our [blog](/blog). ## Frequently asked questions ### Do prediction markets have a liberal or conservative bias? Broad historical studies show no systematic left-right bias in prediction markets over the long term. However, short-term partisan sentiment can cause temporary mispricing, especially on platforms with position limits where enthusiastic retail traders can overwhelm rational arbitrageurs. ### Why do prices for the same event differ across platforms? Prices diverge because different platforms attract different demographics with distinct biases, and structural rules (like position limits or fee structures) prevent arbitrageurs from fully correcting the gaps. Additionally, platforms may have slightly different rules for how an event is officially resolved. ### What is the favorite-longshot bias? This is a well-documented behavioral anomaly where traders tend to overvalue unlikely outcomes (longshots) and undervalue highly probable ones (favorites). This bias is prevalent across many prediction markets, including sports betting and political contracts. ## Sources * [National Bureau of Economic Research: ARE PREDICTION MARKETS POLITICALLY BIASED?](https://www.nber.org/papers/w35846) * [EFMA: Partisan Sentiment and Returns from Online Political Betting Markets in the 2020 U.S. Presidential Election](http://www.efmaefm.org/0EFMAMEETINGS/EFMA%20ANNUAL%20MEETINGS/2025-Greece/papers/Predictit3.pdf) * [CEPR: The economics of the Kalshi prediction market](https://cepr.org/voxeu/columns/economics-kalshi-prediction-market) * [National Bureau of Economic Research: LIMITS TO ARBITRAGE AND PREDICTION MARKET EFFICIENCY: THE CASE OF PREDICTIT](https://www.nber.org/system/files/working_papers/w35845/w35845.pdf) * [American Finance Association: Peering into the Black Box: Trader Strategies in the Iowa Electronic Markets](https://afajof.org/management/viewp.php?n=79184) * [PredictIt: platform announcements](https://www.predictit.org/platform-announcements) * [PredictIt: deposits and withdrawals update](https://www.predictit.org/stories/detail/102/Product-Update--Deposits-and-Withdrawals-are-Free) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) --- ### Prediction Market Arbitrage Capital Lockup and Expected Value Canonical: https://predictionmarketarbitrage.com/blog/expected-value-capital-lockup-arbitrage Published: 2026-11-09 Updated: 2026-11-16 Author: PMA Research Summary: Learn how settlement delays, fees, opportunity cost, and resolution risk change the expected value of a prediction market arbitrage position. Finding a 2% gross edge across two prediction markets feels like a rare victory. But if that capital remains trapped for nine months waiting for an oracle to settle, your brilliant trade might actually be losing money in real terms. Understanding the math behind time value, opportunity cost, and resolution risk separates profitable arbitrageurs from those who merely lock up liquidity for free. ## Key takeaways * A static arbitrage spread must be annualized to reveal its true yield, as a 2% gross return over nine months severely underperforms risk-free cash. * Capital locked in unresolved prediction market contracts incurs an opportunity cost directly tied to current short-term Treasury bill rates. * Yield-bearing collateral, such as the variable interest sometimes paid on eligible exchange balances, can partially offset the drag of long-dated lockups. * Sizing long-dated trades requires strict Kelly-style caution, because tying up your bankroll prevents you from capturing higher-yield, short-term opportunities. ## What is the true cost of capital lockup? When you execute a cross-venue arbitrage, you buy a YES contract on one exchange and a NO contract on another, aiming for a $1.00 payout at resolution. The math seems simple: if the combined cost is 98 cents, you lock in a 2-cent gross profit. However, prediction markets are fully collateralized. Your 98 cents sits in escrow until the market resolves. This is the fundamental friction of collateralized prediction markets. Every dollar of open interest is a dollar of idle cash. If the event resolves in nine months, that 2% gross return is your total yield for the period. Annualized, a 2% return over 270 days is roughly 2.7%. This figure ignores the friction of execution. You must account for trading costs using a [fee calculator](/fee-calculator). For instance, the Kalshi general taker fee formula is 0.07 multiplied by the contract multiplier, the number of contracts, the price, and one minus the price. The fee and position cost are rounded up to a centicent, though printed fee tables often display rounding to cents. At a default multiplier of 1, this peaks at approximately 1.75 cents per contract at 50 cents, but current precision must always be read from live fee parameters, as multipliers and exceptions vary. Polymarket applies taker fees to most categories (e.g., 0.07 for Crypto, 0.04 for Politics, and 0 for Geopolitics), calculated to exactly five decimal places in USDC, not cents. Makers on Polymarket are never charged fees and can earn rebates, though rebate eligibility is not guaranteed revenue. Once fees are deducted, your 2.7% annualized gross return easily shrinks below 2%. Furthermore, operational hazards exist; malicious actors distribute software claiming to automate these trades, making a [fake bot checklist](/fake-bot-checklist) essential to ensure you do not lose your private keys. The true cost of capital lockup is the combination of these fees, the time delay, and the inability to pivot your strategy. ## How do Treasury yields compare to prediction market returns? To evaluate an arbitrage opportunity, you must compare its annualized net return against the risk-free rate. Idle cash has an opportunity cost. If you park your funds in a government bond, you earn a yield without platform risk or oracle disputes. As of November 2026, the four-week Treasury bill secondary market rate sat at 3.89% [FRED: 4-Week Treasury Bill Secondary Market Rate](https://fred.stlouisfed.org/series/DTB4WK). If your nine-month arbitrage trade yields an annualized 2% after fees, you are actively choosing to underperform a standard T-bill. You are taking on smart contract risk, exchange counterparty risk, and the possibility that venues resolve the same event differently, all to earn less than you would in a standard brokerage sweep account. A below-par price on a near-certain contract is often just a reflection of this maturity-dependent settlement discount [[Working Paper] arXiv: When Certainty Is Not Worth It: Capital Lock-Up and Settlement Discounting in Prediction Markets](https://arxiv.org/pdf/2605.31431). ## Does interest on balances offset the lockup penalty? The opportunity cost equation changes if the exchange pays you to keep your capital on the platform. Yield-bearing collateral allows long-dated prices to sit closer to their true probabilities because the carry cost drops. Some regulated exchanges, such as Kalshi, have offered variable APY on cash and open positions under specific promotional or balance conditions. If your capital is earning a baseline yield while locked in a nine-month arbitrage trade, the gross spread becomes a bonus on top of the baseline yield, rather than your sole source of return. However, these rates are variable, subject to change, and are never guaranteed. ## How to calculate expected value with resolution risk? Expected value (EV) in cross-venue arbitrage is rarely a pure risk-free calculation. You must price in resolution risk. Venues can and do resolve the same real-world event differently based on their specific rulebooks. For example, Polymarket relies on the UMA optimistic oracle (where disputes can take days) and Chainlink TWAP for up-down markets, while Kalshi uses its own source-based rulebook. If a dispute delays settlement, your annualized return plummets. If a rules committee rules against your interpretation on one leg, you could lose your entire principal. Consider a hypothetical $10,000 arbitrage trade resolving in nine months. You find a 2% gross spread. After fees, your net profit is $150 (1.5%). You estimate a 1% chance that the venues resolve differently, causing both legs to settle at $0.00. ### Expected Value Comparison Table | Scenario | Probability | Payout | Net Profit/Loss | EV Contribution | | :--- | :--- | :--- | :--- | :--- | | Normal Resolution | 99.0% | $10,000 | +$150 | +$148.50 | | Mismatched Resolution | 1.0% | $0 | -$9,850 | -$98.50 | | **Total Expected Value** | **100.0%** | | | **+$50.00** | Your $150 locked profit is actually a $50 expected value once resolution risk is priced in. A $50 return on $9,850 of capital over nine months is an annualized yield of roughly 0.67%. Compared to standard T-bill rates, this hypothetical trade destroys wealth. This is strictly an educational illustration and not financial advice. ## Why does Kelly sizing caution against long-dated arbitrage? Even if you find a long-dated arbitrage opportunity that beats the risk-free rate after adjusting for fees and resolution risk, you must consider bankroll management. The Kelly criterion and similar sizing models dictate how much of your portfolio to allocate to a specific edge to maximize long-term growth. A core assumption of these models is that capital can be recycled. In prediction markets, capital lockup breaks this assumption. Tying up 20% of your bankroll in a nine-month trade prevents you from deploying that same capital into a 0.5% edge that resolves tomorrow. Short-duration trades can be compounded dozens of times over a nine-month period. When you commit funds to a long-dated lockup, you are paying a massive, hidden opportunity cost in the form of missed future trades. Strict sizing caution is required; long-dated positions should only command a fraction of the capital you would allocate to a short-term trade with the exact same expected value. Professional arbitrageurs often demand a significantly higher hurdle rate for long-dated trades to compensate for this loss of liquidity. If a trade does not clear this elevated hurdle rate, the capital is better left in cash or deployed into rapid-turnover markets. ## Frequently asked questions ### What is capital lockup in prediction markets? Capital lockup occurs because prediction markets are fully collateralized. When you purchase a contract, your funds are held in escrow until the market officially resolves, preventing you from using that money for other investments or trades. ### How do taker fees impact arbitrage returns? Taker fees directly reduce your gross spread. For example, Kalshi charges a fee based on the number of contracts, the price, and a multiplier, which at a standard multiplier of 1 peaks at approximately 1.75 cents per contract at 50 cents. Polymarket charges category-specific rates calculated to five decimal places in USDC. Both can easily consume a narrow arbitrage margin. ### Why is annualized return important for arbitrage? Annualized return standardizes the profit over a one-year period, allowing you to compare the trade against baseline investments like Treasury bills. A 2% return over a few days is excellent, but the same return over nine months is highly inefficient. ## Sources * [FRED: 4-Week Treasury Bill Secondary Market Rate](https://fred.stlouisfed.org/series/DTB4WK) * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) --- ### Prediction Market Arbitrage: Scheduled Events vs Surprises Canonical: https://predictionmarketarbitrage.com/blog/scheduled-events-vs-surprises-mispricings Published: 2026-11-09 Updated: 2026-11-16 Author: PMA Research Summary: Compare scheduled economic releases with unexpected news, including liquidity changes, latency risk, and the costs of hypothetical prediction market trades. Prediction markets price scheduled economic releases and sudden surprises very differently, creating distinct structural environments for traders. While retail participants often flock to CPI prints and FOMC days hoping to catch a quick breakout, the reality of latency arbitrage makes trading the exact release minute a losing game. Smart arbitrageurs instead focus on pre-positioned resting orders and structural mispricings that emerge well before the data hits the wire. ## Key takeaways * Scheduled macroeconomic releases force market makers to widen spreads immediately before the print to protect against latency arbitrage. * Retail traders cannot win speed races against institutional bots during the exact minute a scheduled data point goes live. * Pre-positioning non-marketable resting orders allows traders to capture liquidity premiums without competing on millisecond execution speeds. * Sudden, unscheduled surprises often exhibit informational inertia, taking minutes rather than milliseconds to fully price into decentralized venues. * Fee structures vary significantly by venue and market category, requiring precise calculation of multipliers and rounding rules. ## How do prediction markets price scheduled economic data? Scheduled catalysts like the Consumer Price Index (CPI) releases from the Bureau of Labor Statistics or Federal Open Market Committee (FOMC) decisions operate on a strict, public calendar. On platforms like Kalshi, these events trade as a ladder of mutually exclusive outcomes. For example, a CPI market might feature nested strikes asking if the month-over-month inflation will print above specific percentage thresholds. Because the exact time of the release is known, the market behavior follows a predictable lifecycle. In the days leading up to the event, prices aggregate public forecasts and form an implied probability distribution. A working paper analyzing Kalshi data suggests that these implied distributions are broadly consistent with established financial instruments, often allocating probability mass in ways that reflect plausible macroeconomic outcomes [Federal Reserve: Kalshi and the Rise of Macro Markets](https://www.federalreserve.gov/econres/feds/files/2026010pap.pdf). However, as the clock ticks closer to 08:30 ET for a CPI release, the market microstructure changes drastically. Liquidity providers pull their quotes to avoid being picked off by faster participants who receive the data feeds fractions of a second earlier. ## Why is a CPI-release latency race difficult? Attempting to trade the exact moment a scheduled number is published is a trap for retail participants. Institutional trading desks invest heavily in microwave towers and direct fiber lines to receive government data feeds in microseconds. If you are clicking a button on a web interface after seeing a headline on a financial news terminal, you are already the liquidity for someone else's exit. This dynamic was starkly illustrated during the May 2024 CPI leak. The Bureau of Labor Statistics inadvertently uploaded the CPI report to a public server thirty minutes early. A study of this event revealed a massive monitoring infrastructure gap: institutional CME Fed Funds futures adjusted within seconds of the inadvertent upload, while decentralized prediction markets exhibited near-zero top-of-book response for 35 minutes [ScienceDirect: Informational inertia in a decentralized prediction market](https://www.sciencedirect.com/science/article/abs/pii/S1544612326009062). If you try to race the machines on a scheduled release, you will likely lose. For developers attempting to build trading bots, understanding the API infrastructure is critical. On Kalshi, public REST market data requires no authentication, while WebSocket connections require authentication. The REST current example root is `https://external-api.kalshi.com/trade-api/v2`. When querying the orderbook, the `yes_dollars` and `no_dollars` arrays contain dollar prices and fixed-point sizes, sorted ascending with the best bid last. Crucially, only bids are returned; the YES ask is calculated as 1 minus the best NO bid. Because REST market responses are cached and can lag by seconds, they are unsuitable for latency-sensitive release trading. ## What are the mechanics of spread widening before a scheduled catalyst? Spread widening is a defensive mechanism. When a market maker knows a binary resolution is imminent, the risk of holding a stale quote approaches 100 percent. Consequently, the bid-ask spread on a contract trading at 50 cents might widen from 48-52 cents to 30-70 cents in the final minute before an FOMC announcement. This widening reflects the resolution of uncertainty. An NBER working paper on high-frequency event contracts notes that monetary policy announcements significantly impact market-implied macroeconomic expectations [NBER: The Effects of Monetary Policy on Macroeconomic Expectations](https://www.nber.org/system/files/working_papers/w35968/w35968.pdf). Furthermore, the variance of interest rates declines notably following inflation releases, consistent with a general resolution of uncertainty. For traders, this means that entering a position at market price seconds before a release often results in paying a massive premium. ## What risks remain with pre-positioned resting orders? Instead of fighting latency wars, successful arbitrageurs use pre-positioned resting orders. By placing non-marketable limit orders well outside the current spread hours or days before the event, you can act as a liquidity provider when volatility spikes. Note that a limit order can still be a taker if it is marketable upon entry; only resting, non-marketable orders provide liquidity. Consider the following timeline for a hypothetical 08:30 ET CPI release: | Time to Release | Market Phase | Illustrative Risk Review | | :--- | :--- | :--- | | T minus 24 hours | Price discovery | Analyze implied distributions against consensus. | | T minus 2 hours | Check available liquidity | Review resting orders and inventory exposure. | | T minus 5 minutes | Spreads widen | Cancel unneeded orders; do not cross the spread. | | 08:30 ET, if scheduled | Release | Check fills, feed health, and exposure without assuming counterparty motives. | | T plus 10 minutes | Price settles | Evaluate fills and manage positions. | This approach avoids the release minute entirely. It capitalizes on the fact that sudden shocks, especially unscheduled surprises, can cause temporary mispricings. A working paper presented at the European Central Bank suggests that contractionary shocks associated with FOMC statements have statistically significant impacts on the distribution of the Federal Funds rate, often increasing the second moment of uncertainty [European Central Bank: Kalshi and the Rise of Macro Markets](https://www.ecb.europa.eu/press/conferences/shared/pdf/20261005_monpolconf/Wright_paper.pdf). By resting orders, you catch the overreactions. ## Calculating fees and expected value in macro markets When trading these events, understanding the fee structure is critical. A winning binary contract usually pays $1.00 at resolution, though some structures include ranges, partial payouts, or rare 50-50 unknown redeems at $0.50. On Kalshi, the general taker fee formula is `round up(M * 0.07 * C * P * (1 - P))`, where M is the multiplier (default taker M=1), C is the number of contracts, and P is the price. The July 2026 effective schedule states that the fee plus position cost is rounded up to a centicent, though printed fee tables round to cents. Current precision must be read from live fee parameters; do not assume a per-contract 1.75-cent hard cap after rounding or across multipliers. For example, the general fee for 100 contracts at $0.51 is an unrounded $1.7493, which the table displays as about $1.75. Meanwhile, Polymarket charges taker fees based on the category: Crypto is 0.07, Sports/Economics/Culture/Weather/Other is 0.05, Finance/Politics/Mentions/Tech is 0.04, and Geopolitics is 0. Combos have a 0.05 rate. Polymarket fees are rounded to five decimal places in USDC, not cents. For example, 100 shares at a 0.46 price with a 0.05 rate equals exactly 1.24200 USDC. Let us look at a worked numeric example for a hypothetical Kalshi CPI contract priced at 40 cents. If you buy 1,000 YES contracts as a taker (assuming M=1): * Base cost: 1,000 x $0.40 = $400.00 * Unrounded fee calculation: 1 * 0.07 * 1,000 * 0.40 * (1 - 0.40) = $16.80 * Total capital outlay: $416.80 (subject to centicent rounding rules) * Payout if YES: $1,000.00 * Net profit if YES: $583.20 If you instead use resting, non-marketable limit orders (maker), you often avoid the taker fee. On Polymarket, makers are never charged. On Kalshi, the default maker multiplier is M=0, but there are many series exceptions where maker fees apply (calculated as `M * 0.0175 * C * P * (1 - P)`). There is no universal blanket maker-free claim. ## Resolution mechanics and regulatory considerations Venues resolve the same event differently based on their specific rulebooks. Polymarket uses the UMA optimistic oracle for prediction markets and Chainlink TWAP for up-down markets. The rules define the resolution, not the market title. UMA resolutions involve a 2-hour challenge period, and disputes can take days to resolve. Additionally, traders must consider regulatory implications. CFTC regulation is not FDIC insurance, nor does it guarantee risk-free capital. ## Frequently asked questions ### Why do prediction market prices freeze right before a CPI release? Market makers pull their liquidity to avoid being picked off by faster institutional traders who receive the data feeds microseconds before the public. This causes the bid-ask spread to widen so much that trading effectively halts until the new information is fully digested. ### Can I use an API bot to trade the exact minute of an FOMC decision? A release-time strategy requires evidence about feed delays, order latency, available depth, and adverse selection. Retail access alone does not establish either profitability or inevitable loss. REST snapshots can become stale between reads, and streaming data does not guarantee successful execution. ### Do all prediction markets resolve economic data the same way? No. Venues can resolve the same event differently depending on their specific contract rules. One platform might use the unrevised initial print, while another might wait for a specific official publication. Resolution mechanisms also differ, such as Polymarket's use of the UMA optimistic oracle versus centralized settlement. ## Sources * [Federal Reserve: Kalshi and the Rise of Macro Markets](https://www.federalreserve.gov/econres/feds/files/2026010pap.pdf) * [ScienceDirect: Informational inertia in a decentralized prediction market](https://www.sciencedirect.com/science/article/abs/pii/S1544612326009062) * [NBER: The Effects of Monetary Policy on Macroeconomic Expectations](https://www.nber.org/system/files/working_papers/w35968/w35968.pdf) * [European Central Bank: Kalshi and the Rise of Macro Markets](https://www.ecb.europa.eu/press/conferences/shared/pdf/20261005_monpolconf/Wright_paper.pdf) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) --- ### Reading Order Book Depth: Arbitrage Liquidity Limits Canonical: https://predictionmarketarbitrage.com/blog/reading-order-book-depth-liquidity Published: 2026-11-08 Updated: 2026-11-16 Author: PMA Research Summary: Discover how order book depth and liquidity cap your prediction market arbitrage size. Learn to calculate volume-weighted average fill prices and account for specific venue fees. Prediction market arbitrage looks deceptively simple when you only check the top of the order book. In reality, available liquidity dictates exactly how many contracts you can trade before slippage erases your edge. Understanding depth is the only way to calculate your true maximum profitable size. ## Key takeaways * Top-of-book prices are illusions for large orders because slippage increases your average cost as you consume deeper liquidity levels. * The maximum profitable size of an arbitrage trade is capped by the thinner of the two order books, not the venue with the wider headline spread. * A working paper suggests prediction market depth is distributed across multiple price levels rather than concentrated entirely at the best bid or ask. * Calculating your net edge requires factoring in volume-weighted average fill prices alongside specific venue taker fees. ## What does order book depth actually mean in prediction markets? Order book depth represents the cumulative number of contracts available at various price levels. When you look at a market, the quoted price only applies to the specific quantity resting at that exact level. If a YES contract is offered at 52 cents for 40 contracts, the 41st contract will cost you more. In traditional finance, continuous market makers constantly replenish liquidity. Prediction markets operate differently. Most events trade on a single order book with thinner liquidity. A system that only checks the top-of-book price without verifying the depth behind it will report phantom profits. You must measure cumulative notional size within a fixed band to understand what an order will actually cost. Research in the form of working papers suggests that liquidity is layered. A study of decentralized prediction markets found that the top-of-book carries a median top-level share of about 13.6% of top-ten-level depth in its sample, leaving about 86.4% at the other nine levels [arXiv: The Anatomy of a Decentralized Prediction Market](https://arxiv.org/html/2604.24366). That sample-specific finding is a reason to inspect deeper levels, not a universal depth profile for every venue or period. ## How do you calculate volume-weighted average fill price? To find your true execution cost, you must walk the book. This means consuming each price level in turn until your target size is filled, then computing the volume-weighted average price. Consider a hypothetical arbitrage trade where you want to buy 1,000 YES contracts on Polymarket and 1,000 NO contracts on Kalshi. A standard binary YES contract typically pays $1.00 at resolution. ### Polymarket YES Ask Book | Level | Price | Quantity at Level | Cumulative Quantity | | :--- | :--- | :--- | :--- | | 1 | $0.40 | 200 | 200 | | 2 | $0.41 | 300 | 500 | | 3 | $0.42 | 400 | 900 | | 4 | $0.43 | 100 | 1,000 | | 5 | $0.45 | 500 | 1,500 | ### Kalshi NO Ask Book | Level | Price | Quantity at Level | Cumulative Quantity | | :--- | :--- | :--- | :--- | | 1 | $0.55 | 100 | 100 | | 2 | $0.56 | 400 | 500 | | 3 | $0.57 | 300 | 800 | | 4 | $0.58 | 200 | 1,000 | | 5 | $0.59 | 600 | 1,600 | If you only looked at the top of the book, you would see YES at $0.40 and NO at $0.55, totaling $0.95. This implies a 5-cent gross profit per contract pair. However, walking the book for 1,000 contracts changes the math. On Polymarket, you buy 200 at $0.40, 300 at $0.41, 400 at $0.42, and 100 at $0.43. Your total cost is $414.00, making your volume-weighted average fill price $0.414. On Kalshi, you buy 100 at $0.55, 400 at $0.56, 300 at $0.57, and 200 at $0.58. Your total cost is $566.00, making your volume-weighted average fill price $0.566. Your combined execution cost is $0.98 per pair. The gross edge has shrunk from 5 cents to 2 cents before fees. This mechanical exercise proves why you must size your expectations to the book rather than the last traded price. ## When does the marginal gap after fees hit zero? Gross edge is only the first hurdle. You must also account for venue fees to find the maximum profitable size. The marginal gap hits zero when the cost of consuming the next level of liquidity, plus the associated fees, equals the $1.00 payout. Fee structures vary significantly. Kalshi charges a general taker fee calculated as M * 0.07 * contracts * price * (1 - price) before venue rounding, where M is a multiplier (default 1), and the fee plus position cost is rounded up to a centicent. This maxes out at about 1.75 cents per contract at the 50-cent mark for standard markets. Polymarket expanded taker fees to nearly every category on March 30, 2026, calculating fees as contracts * feeRate * price * (1 - price) rounded to five decimal places USDC, while makers pay no fee and can earn rebates. Using our previous 1,000-contract example, the gross cost was $0.98. Computing fees separately at every illustrated fill level gives a raw Kalshi fee of $17.1892 and a Polymarket fee of $9.70080 at a hypothetical politics rate of 0.04. The $980 purchase cost plus $26.8900 in fees exceeds the $1,000 matched payout by $6.8900 before Kalshi venue rounding. Under these assumptions the 1,000-pair trade is loss-making even without a settlement mismatch. To find the optimal size, you must iteratively calculate the net edge at increasing quantities. You stop adding size the moment the next available pair costs more than $1.00 after fees. You can use our [fee calculator](/fee-calculator) to model these exact thresholds. Always remember that venues can resolve the "same" event differently, adding fundamental risk to the execution risk. ## Are thin markets more vulnerable to spoofing and fleeting quotes? A wall of liquidity you can see is a wall that can be canceled for free. Large resting orders can disappear before execution, for legitimate risk-management reasons or potentially abusive ones. Spoofing involves placing large orders with no intention of executing them. A short-lived or canceled quote alone does not establish spoofing: the abusive-intent element cannot be inferred merely from a public snapshot. Treat displayed depth as conditional availability rather than proof of manipulation. To protect yourself, you must track how long quantity remains available and whether the book replenishes after trades. A liquid-looking top level is not the same as a liquid execution path. If you are automating your execution, ensure your logic accounts for partial fills and rejected orders. Review our [fake bot checklist](/fake-bot-checklist) to ensure your execution algorithms are not falling for phantom liquidity. ## How does liquidity change by time of day and market cycle? Depth is not static. It fluctuates based on the time of day, the proximity to resolution, and the broader news cycle. Markets are typically thinnest right after a major news event breaks. Market makers pull their quotes to avoid adverse selection, leaving momentary gaps in the book. While spreads may look massive during these windows, the available depth is usually too small to support a meaningful arbitrage position. Conversely, liquidity may change as an event approaches resolution, but the direction and magnitude depend on the market and other factors. Capital commits as the timeline shortens, thickening the order book across multiple levels. However, long-tail events outside of major political races or macroeconomic data releases often suffer from chronic shallowness. A working paper analyzing decentralized order books suggests that depth decays gently across the first ten levels, but the absolute number of contracts available depends heavily on the market's overall attention [arXiv: The Anatomy of a Decentralized Prediction Market](https://arxiv.org/html/2604.24366v2). Sizing your positions around this reality is the only way to survive. This information is for educational purposes and is not financial advice. ## Frequently asked questions ### What is slippage in a prediction market? Slippage is the difference between the expected price of a trade and the actual price at which it executes. It occurs when your order size exceeds the quantity available at the best quoted price, forcing you to consume deeper, more expensive levels of the order book. ### Why do wide spreads not always mean high profits? A wide spread with no depth behind it is a small opportunity with a large percentage attached. If a 4 percent spread only has $50 of available liquidity, your maximum gross profit is capped at $2 regardless of how much capital you have to deploy. ### How do taker fees impact arbitrage sizing? Taker fees directly reduce your net edge per contract. Because slippage increases your gross cost as you buy more contracts, adding percentage-based or fixed taker fees will cause your marginal profit to hit zero much faster, strictly limiting your maximum trade size. ## Sources * [arXiv: The Anatomy of a Decentralized Prediction Market](https://arxiv.org/html/2604.24366) * [arXiv: The Anatomy of a Decentralized Prediction Market](https://arxiv.org/html/2604.24366v2) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) --- ### How to Spot a Live Prediction Market Mispricing in Real Time Canonical: https://predictionmarketarbitrage.com/blog/how-to-spot-live-prediction-market-mispricing Published: 2026-11-07 Updated: 2026-11-16 Author: PMA Research Summary: Learn how to spot a live prediction market mispricing using order-book depth checks, cross-venue price boards, and strict alert thresholds. Spotting a live prediction market mispricing requires looking past the illusion of the midpoint to the reality of the order book. While basic screeners broadcast theoretical spreads across venues, executable arbitrage only exists when depth, fees, and settlement rules align perfectly. Building a rigorous market-watch method separates phantom gaps from genuine inefficiencies. ## Key takeaways * Midpoints are deceptive; always measure mispricing against executable bids and asks at your specific target size. * Cross-venue arbitrage requires identical resolution criteria, as seemingly identical markets often settle on different sources or timelines. * Taker fees and slippage consume the majority of theoretical spreads, making depth-adjusted calculations mandatory before execution. * Stale quotes on thin order books create false signals that vanish the moment a taker order is routed to the matching engine. ## Why do midpoints create phantom mispricing signals? A prediction market is mispriced when the price you can actually trade at differs from the true probability, not when the midpoint looks off. Many traders rely on basic screeners that rank opportunities by midpoint gaps. This approach fails because a one-sided book cannot be mispriced in any usable sense. Cheap longshots and near certainties frequently lack bids entirely. On these markets, the displayed midpoint is computed against nothing. Missing asks are unavailable, never zero. When you see a contract priced at 3 cents that you believe is worth 1 cent, it might look like a screaming short. However, at low prices, the tick size, the fee, and the missing bid become massive relative to the price itself. A wide spread does not mean the market is wrong; it means you must cross a massive percentage threshold just to find out. Relying on midpoints ignores the mechanics of execution, which is why taker fees frequently eliminate theoretical arbitrage before a trade can even be placed. Remember that a limit order can be a taker if it is marketable; only resting, non-marketable orders provide liquidity. ## How to construct a cross-venue price board? To spot real inefficiencies, you need a cross-venue price board that ingests streaming data. For example, the Kalshi REST API (currently rooted at `https://external-api.kalshi.com/trade-api/v2`) provides public market data without authentication, but these REST responses are cached and can lag by seconds. Polling introduces latency that erases the edge before you can act on it. Streaming ingestion pulls live order-book and trade data over authenticated WebSocket connections, ensuring your board reflects the current state of the matching engine. Your price board must implement canonical ID mapping. This assigns a single internal identifier to each contract regardless of venue-specific naming conventions. Normalization then converts each venue's price convention, fee schedule, and settlement rule into a common schema. Time alignment stamps every event to a shared clock, correcting for venue-side clock drift. Deduplication drops repeated snapshots so smoothing calculations do not double-count static periods. Without this infrastructure, you are comparing apples to oranges. A robust price board highlights transient spread widening around news events, when market makers pull quotes and bid-ask spreads blow out on one venue while the other stays tight. ## What distinguishes real mispricing from structural differences? Genuine mispricing is rare. Most apparent gaps reflect different resolution wording, different fee structures, or different access regimes. Venues can resolve the "same" event differently. For instance, Polymarket relies on the UMA optimistic oracle for prediction markets (where disputes can take days) and Chainlink TWAP for up-down markets. If you trade a gap across venues with different oracles, you are not hedged. You hold two different directional bets. Fees also distort the picture. While a standard winning token usually pays $1.00 at resolution, this is not universal (for example, rare Unknown/50-50 outcomes on Polymarket redeem at $0.50 each). You must calculate exact fees. The Kalshi general taker fee formula is `M * 0.07 * C * P * (1 - P)` before venue rounding, where the default taker multiplier M=1. The July 2026 schedule states that the fee plus position cost is rounded up to a centicent, though printed tables round to cents. You must read current precision from live fee parameters and not assume a per-contract 1.75-cent hard cap across multipliers [Fee Schedule for July 2026 - 7.7.26 Update](https://kalshi.com/docs/kalshi-fee-schedule.pdf). Meanwhile, Polymarket calculates fees using `fee = C * feeRate * p * (1 - p)`. Rates vary by category: Crypto is 0.07, Sports is 0.05, and Geopolitics is 0. Makers are never charged. Crucially, Polymarket fees are rounded to five decimal places in USDC, not cents [Fees](https://docs.polymarket.com/trading/fees). If you do not account for these exact structures, your perceived edge is an illusion. ## How to check order-book depth and stale quotes? The price you see on a market summary screen is just one number describing a structure that has more important numbers buried inside it. The order book is where the real story lives. A gap is only theoretical when the displayed prices do not support your required size. To verify an opportunity, you must walk the book. This means measuring depth at every price level required for your planned size. You must calculate the depth-adjusted average fill price, applying slippage first and fees second. Walking both legs' ask ladders in lockstep is the only way to find the largest hedge that still clears the viability gate. Both directions must use actual buy asks. Below is a hypothetical market-watch log table illustrating how to track these variables (note that return calculations must include both legs' capital and costs): The following scan log is entirely hypothetical. Depth is shown in contracts and the target is 100 pairs. The calculation assumes a Kalshi taker multiplier of 1, a Polymarket coefficient of 0.04, and no other costs. Net margin uses the raw Kalshi formula before venue rounding; displayed liquidity is not a fill guarantee. | Illustrative Time | Kalshi YES Ask | Polymarket NO Ask | Kalshi Depth | Polymarket Depth | Modeled Net per Pair | Review | | :--- | :--- | :--- | :--- | :--- | :--- | :--- | | 10:04:12 | $0.43 | $0.53 | 450 | 120 | +$0.012879 | Candidate, verify execution | | 10:04:15 | $0.43 | $0.54 | 450 | 15 | +$0.002907 | Reject at target size: insufficient depth | | 10:05:00 | $0.41 | $0.53 | 800 | 600 | +$0.033103 | Candidate, verify execution | | 10:12:30 | $0.49 | $0.50 | 100 | 500 | -$0.017493 | Reject: modeled fees exceed gap | This log captures the exact moment a trade becomes viable and the exact reason it fails. Logging every rejected match with its failure reason is the fastest way to see which fields cause the most false positives. ## What alert thresholds actually work for live logging? A production detector needs strict alert thresholds to prevent alert fatigue. Rank alerts by conditioned surprise rather than raw gap size. Weight each signal by inverse liquidity and treat trade size as a primary feature. Your alert logic must require live real asks, confirming the bot has prices it can actually attempt to buy against. Both markets must be open, as a halted leg turns a hedge into a naked position. The spread must survive the actual applicable cost model, and the markets need to represent compatible settlement outcomes. Remember that Fill-or-Kill (FOK) orders control fills on one venue only; they are never atomic cross-venue. Be highly skeptical of automated tools promising guaranteed returns. Academic working papers, such as recent preprints on combinatorial arbitrage [Combinatorial Arbitrage in Prediction Markets](https://arxiv.org/html/2610.23517v2), highlight theoretical opportunities but are not peer-reviewed and often assume ideal execution conditions. Independent logging helps validate a scanner, but neither custom software nor third-party tools can ensure capital security or profitable fills. This is not financial advice, but a structural reality of modern prediction markets. ## Frequently asked questions ### What causes market divergence signals across prediction market venues? Market divergence signals appear when correlated contracts on different venues price the same event differently. This usually happens because one venue's order book absorbs new information faster than the other, or due to structural differences between an on-chain venue and a regulated order book. ### How do I avoid false divergence signals from resolution mismatches? You must confirm the outcome definition, resolution cutoff, timezone, and settlement source before treating two contracts as the same market. A gap driven by a settlement mismatch is uncompensated risk, not arbitrage. ### Can settlement rules break a prediction-market arbitrage? Yes. A trade can look matched economically and still fail to offset if the venues use different sources, deadlines, or dispute processes. Settlement delays can also tie up capital longer than expected, altering the trade's real return. ## Sources * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) * [Combinatorial Arbitrage in Prediction Markets (Working Paper)](https://arxiv.org/html/2610.23517v2) --- ### How Are Prediction Market Profits Taxed? Arbitrage Guide Canonical: https://predictionmarketarbitrage.com/blog/prediction-market-profits-tax-treatment Published: 2026-11-06 Updated: 2026-11-16 Author: PMA Research Summary: Prediction market taxes remain uncertain. Learn how the IRS treats event contracts, Form 1099-DA rules, and record-keeping for arbitrageurs. Prediction market profits raise two separate questions: how much money the trade made, and how tax law classifies that result. A hedged position can have a small economic profit while producing gains and losses on different platforms, in different years, or under different tax rules. This guide explains the reporting issues a US individual should discuss with a qualified tax professional. It does not assign one tax treatment to every event contract or provide personal tax advice. ## Key takeaways * A platform's regulatory status does not, by itself, establish federal tax character. * Capital, Section 1256, wagering, and business-income rules have different conditions; taxpayers cannot freely choose whichever produces the lowest bill. * Beginning in 2026, IRS Publication 505 describes a wagering-loss deduction limited to the lesser of 90% of losses or gambling winnings. That matters only if wagering rules apply to the activity. * Keep records for every leg, fee, settlement, and relevant digital-asset transaction, whether or not an information return arrives. ## Why does an economic hedge not settle the tax question? Suppose a trader buys YES on one platform and NO on another. If the contracts genuinely have complementary settlement rules, the combined position can reduce exposure to the event outcome. It does not merge the two legal instruments into one taxable transaction. Different account owners, reporting periods, contract terms, and classifications may affect the result. Start with the actual instrument, not a website's marketing label. Preserve the contract rules, platform agreement, legal entity, and trading records. Ask the adviser to explain the legal basis for classification and how related positions affect recognition and deductions. Being regulated by the CFTC is not a substitute for satisfying the relevant Internal Revenue Code definition. ## What do the main tax frameworks actually mean? The following table summarizes general US individual tax frameworks, not a menu of elections or a conclusion about a particular venue. | Framework, if applicable | General treatment | Important qualification | | :--- | :--- | :--- | | Capital asset | Gains and losses generally pass through Form 8949 and Schedule D | Holding period and capital-loss limitations matter | | Qualifying Section 1256 contract | Generally year-end mark-to-market and 60% long-term / 40% short-term capital treatment | The instrument must satisfy Section 1256; exchange registration alone is insufficient | | Wagering activity | Winnings and allowable losses follow the wagering rules | The 2026 deduction limitation can leave taxable income despite an economic break-even | | Trade or business | Applicable business-income and expense rules require separate analysis | Trading frequency alone does not establish every claimed business treatment | For capital assets, gains generally are short-term when held one year or less and long-term when held more than one year. After netting capital gains and losses, an individual generally can deduct up to $3,000 of remaining net capital loss against other income, or $1,500 if married filing separately, with remaining losses carried forward subject to the rules. IRS Topic 409 explains these general principles. Qualifying Section 1256 contracts have their own reporting and loss provisions. IRS Form 6781 instructions discuss them, including relevant exceptions and elections. Do not assume that every binary payoff is a regulated futures contract, or describe the short-term portion as a special tax rate separate from short-term capital treatment. ## What changed for wagering losses in 2026? IRS Publication 505 states that beginning in 2026 the gambling-loss deduction on Schedule A is limited to the lesser of 90% of gambling losses or gambling winnings. For an individual applying those rules, itemization and proper records matter. This article does not assert that all prediction market trades are wagers for federal tax purposes. Consider a simplified illustration that assumes the wagering rules apply: $10,000 of winnings and $10,000 of losses produce zero economic profit before other costs. Ninety percent of the losses is $9,000. The deduction ceiling in that illustration is therefore $9,000, not $10,000. Other facts and applicable provisions still need professional review. The example illustrates why a pre-tax arbitrage spreadsheet cannot establish after-tax profitability. Do not apply a new-year rule retroactively to an earlier year's return. Check the publication and forms for the tax year being filed, not merely the date a blog article was published. ## What should I do with platform tax forms? Download the tax documents and account statements actually supplied for the relevant year. A platform may report different activities on different forms. Do not assume that an interest statement covers contract trading, or that another trader's reporting experience applies to your account, jurisdiction, or product. An information return helps reconciliation; it does not eliminate the taxpayer's obligation to report correctly. Likewise, the absence of a form does not establish that income is exempt. Compare reported proceeds, basis, fees, and adjustments with your own ledger, and investigate discrepancies rather than counting the same proceeds twice. Avoid categorical claims that a named platform always or never issues a particular form. Products, intermediaries, reporting requirements, and thresholds can change. Confirm the current year's documents directly with the provider and your adviser. ## How do digital assets change record keeping? The IRS treats digital assets as property for US federal tax purposes. Its digital-assets guidance distinguishes simply holding assets from receiving, selling, exchanging, or otherwise disposing of them. Moving assets between wallets you own is generally different from disposing of them, although paying a network fee with digital assets can itself have reporting implications. Keep the acquisition basis, units, timestamps, and US-dollar values of relevant stablecoin or other digital-asset movements. A stablecoin's target price does not remove the need to retain records. Separate the analysis of the funding asset from the event contract; do not assume that one transaction description resolves both questions. Form 1099-DA concerns digital-asset broker reporting within the scope of its rules. Its existence does not mean every prediction platform must issue it for every event-contract transaction, nor does the form determine the underlying contract's tax character. The IRS instructions describe the applicable reporting years, covered transactions, exceptions, and basis-reporting requirements. ## What belongs in a multi-leg trading ledger? For each transaction, record the venue and account, contract identifier, outcome, quantity, execution time, price, currency, fees, and order or transaction reference. Also retain settlement notices, adjustments, funding records, and relevant wallet transaction hashes. A hypothetical 100-pair trade costing $97 before fees and $2.50 in actual recorded fees has a $99.50 economic outlay. If complementary contracts settle for $100 in total, economic profit is $0.50. That arithmetic does not tell you the taxable character or recognition date of either leg. Preserve both records rather than entering only the fifty-cent net result. Estimate fees with our [fee calculator](/fee-calculator), but use actual charged amounts for reconciliation. If the legs settle in different calendar years, ask about timing and any applicable hedging or straddle provisions before assuming immediate netting. ## Frequently asked questions ### Does CFTC regulation automatically give a contract Section 1256 treatment? No. Regulatory status and tax classification are distinct questions. The instrument must satisfy the relevant tax-law requirements. ### Is income tax-free until I withdraw to my bank? Not necessarily. Selling, exchanging, settling, or otherwise disposing of an asset can matter before a bank withdrawal. Review the actual transactions and applicable rules. ### What about traders outside the United States? Residence, citizenship, source of income, local classification, treaties, and withholding can matter. A foreign platform or a gambling label does not universally make profits tax-free. Consult a qualified adviser in the relevant jurisdiction. ## Sources * [IRS: Topic 409, Capital gains and losses](https://www.irs.gov/taxtopics/tc409) * [IRS: Instructions for Form 6781](https://www.irs.gov/instructions/i6781) * [IRS: Publication 505, Tax Withholding and Estimated Tax, 2026 changes](https://www.irs.gov/publications/p505) * [IRS: Digital assets](https://www.irs.gov/filing/digital-assets) * [IRS: Instructions for Form 1099-DA](https://www.irs.gov/instructions/i1099da) --- ### Building a Prediction Market Data Pipeline Canonical: https://predictionmarketarbitrage.com/blog/prediction-market-data-pipeline-order-books Published: 2026-11-06 Updated: 2026-11-16 Author: PMA Research Summary: Learn how to build a robust prediction market data pipeline. We cover WebSocket connections, order book schemas, and storage for backtesting. Capturing tick-level data from prediction markets is the foundation of any serious quantitative strategy. Relying on stale REST API pulls will leave you executing against phantom liquidity, while a robust WebSocket pipeline ensures you see the true state of the limit order book. This guide breaks down the architecture for streaming, validating, and storing market data for rigorous backtesting. ## Key takeaways * Kalshi and Polymarket require distinct WebSocket handling, from authentication handshakes to sequence tracking. * Storing raw order book deltas in Parquet or DuckDB provides a highly compressed, query-optimized format for backtesting. * Sequence gaps in incremental updates mandate immediate snapshot requests to prevent corrupted local order book states. * Accurate timestamp synchronization is critical, as venues process the exact same real-world event at different latencies. ## How do you connect to prediction market WebSockets? Establishing a reliable connection is the first step in building your data pipeline. Different exchanges implement WebSocket protocols with varying authentication and subscription models. For Kalshi, the production endpoint is located at `wss://external-api-ws.kalshi.com/trade-api/ws/v2` as detailed in their [API Environments and Endpoints](https://docs.kalshi.com/getting_started/api_environments) documentation. Unlike some platforms that allow public access without credentials, Kalshi requires authentication headers directly within the initial HTTP handshake, even if you only intend to consume public market data. If the handshake fails, the server returns an HTTP error rather than upgrading the connection, as noted in the [WebSocket Connection](https://docs.kalshi.com/websockets/websocket-connection) guide. For REST API pulls, the current example root is `https://external-api.kalshi.com/trade-api/v2`, which requires no authentication for public market data, though responses are cached and can lag by seconds. Polymarket utilizes a different architecture. Their Central Limit Order Book (CLOB) market channel is accessible via `wss://ws-subscriptions-clob.polymarket.com/ws/market`, which streams public order book, price, and market lifecycle updates according to the [Polymarket API](https://docs.polymarket.com/getting-started/api) documentation. When building your connection handlers, write your own implementation rather than relying on unverified open-source wrappers. Malicious actors frequently target algorithmic traders; in fact, fake Polymarket arbitrage bot repos have stolen private keys. Always review our [fake bot checklist](/fake-bot-checklist) before integrating third-party code. ## What is the best schema for order book and trade data? Normalizing data across multiple venues is mandatory for cross-market analysis. A standard binary YES contract typically pays $1.00 at resolution on both platforms, though rare Unknown/50-50 outcomes on Polymarket can redeem at $0.50. Furthermore, the raw data structures differ significantly. For instance, Kalshi's REST orderbook_fp arrays contain dollar strings, and the best YES ask is derived from the best opposite NO bid. Your internal schema must standardize timestamps, price formats, and side indicators. Below is a recommended schema for storing normalized trade events: | Column Name | Data Type | Description | | :--- | :--- | :--- | | `exchange` | VARCHAR | Venue identifier (e.g., 'kalshi', 'polymarket') | | `market_ticker` | VARCHAR | Standardized market ID | | `timestamp_ms` | BIGINT | Exchange timestamp in milliseconds | | `local_ts_ms` | BIGINT | Local machine timestamp upon receipt | | `side` | VARCHAR | 'yes' or 'no' | | `price` | DECIMAL | Execution price in USD | | `size` | DECIMAL | Number of contracts traded | | `is_snapshot` | BOOLEAN | True if part of an initial book build | When calculating potential profitability in your backtests, you must account for exact fee structures. For example, the general Kalshi taker fee formula is `round up(M * 0.07 * C * P * (1 - P))`, where M is a multiplier (default taker M=1). For 100 contracts at $0.51, the unrounded fee is $1.7493, but precision must be read from live fee parameters as rounding rules apply (e.g., fee + position cost rounded up to a centicent). There is no universal blanket maker-free claim on Kalshi; certain series have exceptions. Conversely, Polymarket charges taker fees across most categories (e.g., 0.07 for Crypto, 0.05 for Sports/Economics/Culture/Weather/Other, 0.04 for Finance/Politics/Mentions/Tech, and 0 for Geopolitics). Makers are never charged fees and may be eligible for rebates, though eligibility does not guarantee revenue. Polymarket fees are rounded to five decimal places in USDC, not cents. If you buy 100 shares at a 0.46 price with a 0.05 rate, the fee is 1.24200 USDC, not exactly 1.24. Your schema should ideally include a `fee_estimate` column to track these costs dynamically. ## How do you handle sequence gaps and reconnects? Maintaining an accurate local order book requires flawless processing of incremental updates. Dropping a single message will corrupt your book state, leading to disastrous trading decisions. Kalshi manages this by sending an initial `orderbook_snapshot` followed by incremental `orderbook_delta` messages. Each message includes a `seq` integer. The [Orderbook Updates](https://docs.kalshi.com/websockets/orderbook-updates) documentation explicitly states that `seq` stays contiguous. If your system detects a gap in the sequence numbers, you must immediately halt trading on that market and request a `get_snapshot` action to rebuild the book. Applying deltas out of order will result in crossed books or phantom liquidity. Polymarket handles state verification differently. Each order book response includes a unique hash representing the current state of the book. You must compare this hash with the previous response's hash to determine whether the book changed between reads, as explained in their [Prices and Order Books](https://docs.polymarket.com/market-data/prices-order-books) documentation. If the hashes mismatch unexpectedly, a full refresh is required. ## How should you store tick data for backtesting? Writing raw JSON strings to a standard relational database like PostgreSQL will quickly exhaust your storage and result in abysmal query performance. For tick-level data, columnar storage is vastly superior. We recommend writing incoming WebSocket messages to memory or a fast message broker like Redis, then flushing them to disk as Parquet files partitioned by date and exchange. DuckDB is an exceptional tool for querying these Parquet files directly. It allows you to run complex SQL aggregations over billions of rows on a standard laptop without the overhead of a dedicated database server. For example, you can query the average spread of a specific market directly from a directory of Parquet files using DuckDB, filtering by the `market_ticker` defined in your normalized schema. This approach keeps your storage costs low while maximizing backtesting speed. ## How do you validate data quality before backtesting? Before running any historical simulations, you must rigorously validate your stored data. The most common silent failure in data pipelines is clock drift. Always record both the exchange-provided timestamp and your local machine timestamp (`local_ts_ms`). By analyzing the delta between these two values, you can identify periods of high network latency or internal processing bottlenecks. If the delta spikes during volatile market events, your backtest fills will be overly optimistic. Furthermore, when mapping markets across exchanges, remember that venues can resolve the "same" event differently. Polymarket uses the UMA optimistic oracle for prediction markets (where disputes can take days) and Chainlink TWAP for up-down markets. Rules define resolution, not the title. Cross-venue opposite labels alone do not guarantee a riskless payout. Always verify the exact resolution rules in the contract specifications. Here is a short Python snippet demonstrating how to format a subscription request for Kalshi order book updates: ```python import json def build_kalshi_subscription(message_id, tickers): return json.dumps({ "id": message_id, "cmd": "subscribe", "params": { "channels": ["orderbook_delta"], "market_tickers": tickers } }) # Example usage payload = build_kalshi_subscription(1, ["KXFUT24-LSV"]) print(payload) ``` This pipeline architecture provides the reliability required for professional quantitative research. Note that this is not financial advice; always test your systems thoroughly in a paper-trading environment before deploying capital. ## Frequently asked questions ### How much storage do I need for prediction market tick data? It depends on the number of markets you track, but storing raw JSON can consume gigabytes per day. By converting the data to Parquet format, you can compress this by 80-90%, making it feasible to store years of tick data on a standard solid-state drive. ### Can I use REST APIs instead of WebSockets for backtesting data? REST APIs are insufficient for building accurate historical order books because they only provide snapshots at the exact moment of your request, and responses are often cached. WebSockets push every incremental change, ensuring you capture fleeting liquidity and transient price spikes that REST polling will miss. ### What is the difference between a trade feed and an order book delta? A trade feed only reports executed transactions, showing where buyers and sellers actually met. Order book deltas show changes in resting limit orders, including cancellations and new quotes that never resulted in a trade, which is crucial for understanding market depth and calculating slippage. Check our [glossary](/glossary) for more definitions. ## Sources * [Kalshi: API Environments and Endpoints](https://docs.kalshi.com/getting_started/api_environments) * [Kalshi: WebSocket Connection - API Documentation](https://docs.kalshi.com/websockets/websocket-connection) * [Polymarket: API](https://docs.polymarket.com/getting-started/api) * [Kalshi: Orderbook Updates](https://docs.kalshi.com/websockets/orderbook-updates) * [Polymarket: Prices and Order Books](https://docs.polymarket.com/market-data/prices-order-books) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) --- ### Resolution Mismatch: How a Prediction Market Hedge Can Fail Canonical: https://predictionmarketarbitrage.com/blog/resolution-mismatch-arbitrage-gone-wrong Published: 2026-11-05 Updated: 2026-11-16 Author: PMA Research Summary: A hypothetical case study shows how different resolution rules can cause both legs of an apparent hedge to lose, with a practical pre-trade checklist. Prediction market arbitrage promises risk-free returns by trading the spread between venues, but this assumes both platforms define reality the same way. When resolution rules diverge, a perfectly hedged position can collapse, leaving traders with losses on both sides of the trade. Understanding these structural mismatches, alongside the exact fee structures of each platform, is the difference between capturing yield and funding someone else's payout. ## Key takeaways * Arbitrageurs can lose both legs of a trade if platforms use different resolution sources, deadlines, or definitions for the exact same real-world event. * Polymarket relies on the UMA optimistic oracle for dispute resolution, which features a 2-hour challenge period and can occasionally diverge from official sources. * Kalshi strictly enforces its CFTC-regulated rulebook, often settling on specific agency data snapshots that may contradict broader media consensus. * Traders must audit contract wording, expiration times, and fallback clauses before executing cross-exchange positions to avoid catastrophic resolution mismatches. * U.S. taxpayers face classification uncertainty; there is no universal gambling or Section 1256 capital classification for all event contracts. ## How do prediction market resolution rules differ? The foundation of any prediction market is its rulebook. While two markets might appear identical on the surface, their underlying settlement criteria often vary significantly. Kalshi binds each contract to a named source agency and frequently utilizes morning expiration times. In contrast, Polymarket often relies on a consensus of credible reporting and typically sets deadlines at midnight. These differences are not merely academic. They dictate who gets paid. Furthermore, trading costs complicate the picture. Polymarket expanded taker fees to nearly every category on 30 March 2026. Makers pay no fee and can earn rebates, though eligibility does not guarantee revenue. Polymarket taker fees are calculated as `fee = C × feeRate × p × (1 - p)`, rounded to five decimal places in USDC. Rates vary by category: Crypto is 0.07; Sports, Economics, Culture, Weather, and Other are 0.05; Finance, Politics, Mentions, and Tech are 0.04; Geopolitics is 0; and Combos are 0.05. Meanwhile, the Kalshi general taker fee formula is `round up(M × 0.07 × C × P × (1 - P))`, where the default multiplier M is 1. The fee plus position cost is rounded up to a centicent. Traders must read current precision from live fee parameters and should not assume a per-contract 1.75-cent hard cap after rounding or across multipliers. If you are calculating margins, you must account for the exact venue fees and the risk that the underlying event resolves differently on each platform. While a standard winning token typically pays $1.00 and a losing token $0.00, this is not universal; rare Unknown or 50-50 outcomes can redeem at $0.50 each, and other structures include ranges or partial payouts. ## What happens when an arbitrage trade loses both legs? The nightmare scenario for any arbitrageur is a dual-leg failure. This occurs when a trader buys YES on one platform and NO on another, expecting one to pay out $1.00, but due to resolution mismatches, both legs resolve as losers. Consider a hypothetical government shutdown market. Platform A defines a shutdown as a lapse in appropriations lasting past 11:00 AM ET, sourced strictly from the Office of Personnel Management. Platform B defines it as any lapse in funding by 11:59 PM ET, sourced from credible media consensus. If a funding gap occurs at midnight but is resolved by 10:00 AM ET, Platform B might resolve YES, while Platform A resolves NO. If you bought NO on Platform B and YES on Platform A, you lose both investments. ### Hypothetical P&L of a failed hedge (100 Contracts) Return calculations must include both capital and costs. The following table illustrates a hypothetical scenario where a trader buys 100 contracts on two different platforms, paying a 0.04 rate on Platform A and a 0.07 rate on Platform B. | Leg | Venue | Position | Capital | Fee Cost | Total Cost | Resolution | Payout | Net Profit | | --- | --- | --- | --- | --- | --- | --- | --- | --- | | 1 | Platform A | YES at $0.45 | $45.00 | $0.99 | $45.99 | NO | $0.00 | -$45.99 | | 2 | Platform B | NO at $0.50 | $50.00 | $1.75 | $51.75 | YES | $0.00 | -$51.75 | | Total | | | $95.00 | $2.74 | $97.74 | | $0.00 | -$97.74 | In this hypothetical scenario, the trader spent $97.74 expecting a guaranteed $100.00 payout, but instead lost the entire principal and fees. This is not financial advice, but a mathematical reality of mismatched rules. ## Why do UMA oracle disputes cause settlement mismatches? Polymarket utilizes the UMA optimistic oracle for decentralized dispute resolution on prediction markets, while using Chainlink TWAP for up-down markets. The UMA system relies on token holders to vote on the outcome of disputed markets after a 2-hour challenge period. Disputes can take days to fully resolve. Because rules define the resolution rather than the market title, this mechanism can introduce discrepancies between the actual event and the adjudicated outcome. A published AFT 2025 conference paper indicates that during the settlement window, prices often forecast the adjudication process rather than the real-world event itself. When one platform uses a decentralized oracle and another uses a centralized compliance team, the same event can easily yield two different answers. Cross-venue opposite labels alone do not guarantee a payout. ## Can ambiguous wording ruin a hedged position? Ambiguous contract wording is a primary driver of resolution failures. When market creators fail to define key terms, platforms are forced to interpret the rules retroactively. Consider a hypothetical market asking if a political figure will wear a suit by a specific date. If the figure wears a blazer and trousers without a tie, one platform's oracle voters might argue the lack of a tie disqualifies it and vote NO, while another platform might resolve it as YES based on general media consensus. Similarly, consider a hypothetical Super Bowl halftime market asking if a specific guest artist will perform. If the artist appears on stage but does not use a microphone, a strictly regulated platform might determine the event is ambiguous under its performance definition and settle the contract at the last traded price, while a consensus-based platform might resolve the identical event as YES. Venues can resolve the same event differently, turning a perceived arbitrage opportunity into a total loss. ## Pre-trade resolution checklist for arbitrageurs To mitigate the risk of resolution mismatches, traders must conduct rigorous due diligence before executing cross-platform trades. Relying on automated tools without manual verification is dangerous; malicious arbitrage bot repositories can steal private keys from unsuspecting users. ### The resolution audit checklist * Verify the exact expiration date, time, and time zone for both contracts. * Identify the primary resolution source named in each rulebook. * Check for fallback clauses or secondary sources if the primary source is unavailable. * Confirm the precise definition of the event (e.g., signed into law versus announced by media). * Review historical dispute outcomes for similar markets on both platforms. * Calculate maximum potential losses, including all venue-specific fees, if both legs resolve against your positions. By systematically verifying these elements, traders can identify structural mismatches before committing capital. ## Frequently asked questions ### What is a resolution mismatch? A resolution mismatch occurs when two prediction markets asking about the same real-world event settle on different outcomes. This usually happens because the platforms use different data sources, deadlines, or definitions in their contract rules. ### How does the UMA oracle affect Polymarket resolutions? The UMA oracle uses token-holder voting to resolve disputed markets on Polymarket after a 2-hour challenge period. Because outcomes are determined by votes based on specific rule interpretations rather than a centralized authority, the final resolution can sometimes diverge from official government data or media consensus. ### Can I lose money on a perfectly hedged arbitrage trade? Yes. If you buy opposing positions on two different platforms and their resolution rules diverge, it is entirely possible for both contracts to resolve as losers, resulting in a total loss of the capital deployed for the trade, plus any trading fees incurred. ### Are prediction market payouts always exactly $1.00? No. While binary fixed payouts are usually $1.00, rare Unknown or 50-50 resolutions can redeem at $0.50 each. Other market structures may include ranges or partial payouts depending on the venue's specific rules. ## Sources * [Polymarket: resolution](https://docs.polymarket.com/polymarket-learn/markets/how-are-markets-resolved) * [Kalshi: market-data documentation](https://docs.kalshi.com/getting_started/quick_start_market_data) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) --- ### PredictIt, ForecastEx & Newer Venues: Arbitrage Potential Canonical: https://predictionmarketarbitrage.com/blog/predictit-manifold-forecastex-arb-potential Published: 2026-11-04 Updated: 2026-11-16 Author: PMA Research Summary: Explore prediction market arbitrage across secondary venues like PredictIt, ForecastEx, and Manifold. Learn which frictions impact your edge. Arbitrage across prediction markets sounds like guaranteed profit until fees, position limits, and mismatched resolution rules consume the margin. While major platforms dominate the volume, secondary venues like PredictIt, ForecastEx, and Robinhood offer distinct pricing structures that either create or destroy cross-exchange edge. Finding realistic pairings requires mapping the exact cost of capital and execution on every leg of the trade. ## Key takeaways * PredictIt's historical $850 downside risk limit (which was raised to $3,500 per contract on July 15, 2025) and 10% fee on profits structurally restrict arbitrage, often leading to overpriced longshots. * ForecastEx charges a flat $0.01 exchange fee per contract via Interactive Brokers, avoiding the variable curves seen on other platforms. * Polymarket charges taker fees across multiple categories, while makers pay no fee and can earn rebates. Fees are rounded to five decimal places in USDC. * Play-money platforms like Manifold offer no direct financial arbitrage but serve as highly sensitive signal generators for real-money execution. ## How do position limits impact PredictIt arbitrage? PredictIt operates under specific regulatory constraints that fundamentally alter its market structure. Historically, a significant friction was the $850 limit on downside risk per contract (raised to $3,500 in July 2025). This cap prevents well-capitalized traders from aggressively selling overpriced outcomes to bring prices back to fair value. According to academic analysis (which is not peer-reviewed), this restriction means a trader betting on a low-probability outcome can take a much larger position than one betting against it, directly contributing to the overpricing of longshots [NBER: LIMITS TO ARBITRAGE AND PREDICTION MARKET EFFICIENCY: THE CASE OF PREDICTIT](https://www.nber.org/system/files/working_papers/w35845/w35845.pdf). Furthermore, PredictIt charges a 10% fee on profits from winning positions. If you attempt to arbitrage a market by buying "No" shares on every candidate in a primary election, the 10% platform fee on profits from winning shares will often erase the theoretical gross margin. While the platform removed its 5% withdrawal fee on September 23, 2026 (making ACH free, with subsequent card deposits at 3% after a qualifying next-deposit waiver up to $1,000), the combination of position limits and the retained 10% profit fee means PredictIt is rarely viable for pure risk-free arbitrage against other venues. It is better utilized as a standalone directional trading platform. ## What are the true costs of trading on ForecastEx and Robinhood? Newer entrants into the regulated event contract space utilize traditional brokerage rails, which changes the fee math. ForecastEx, accessed via Interactive Brokers, operates on a flat fee model. Interactive Brokers lists a $0.00 commission for ForecastEx contracts, passing through a $0.01 third-party exchange fee per contract [Interactive Brokers: Prediction Markets](https://www.interactivebrokers.com/en/pricing/commissions-events.php). Because a YES contract pays $1.00 at resolution, a single cent represents exactly 1% of the payout regardless of the entry price. Robinhood routes event contracts through its derivatives arm, applying a capped commission structure. A commission cap is not necessarily the total charge: Robinhood documents its commission separately from exchange fees, and rates can depend on the product and account terms. Check the [current event-contract overview](https://robinhood.com/us/en/support/articles/event-contracts-overview/) for the actual route before comparing costs. One critical mechanical difference with ForecastEx is that instruments cannot be sold directly. To exit or reduce a position, a trader must buy the opposing event contract at the same strike, and the broker nets the opposing positions together automatically [Interactive Brokers: Event Contracts in the Web API](https://www.interactivebrokers.com/campus/ibkr-api-page/event-contracts/). The opposing purchase can require additional available funds before netting; the amount depends on the prices, fees, and broker handling, not a universal doubling of capital. Include those requirements in any [fee calculator](/fee-calculator) model. ## Can you arbitrage Polymarket against Kalshi or Novig? Pairing the largest crypto prediction market against regulated US exchanges is the most common arbitrage strategy, but fee structures are complex. Polymarket charges taker fees across multiple categories, while makers pay no fee and can earn rebates. The taker fee rate varies by category, such as 0.07 for Crypto and 0.04 for Politics [Polymarket: Fees](https://docs.polymarket.com/trading/fees). Kalshi uses a variable taker fee formula: `round up(M * 0.07 * C * P * (1 - P))`. This means the fee peaks at exactly 50 cents, costing a maximum of about 1.75 cents per contract, and decreases symmetrically toward the extremes of 1 cent and 99 cents. Current precision must be read from live fee parameters. ### Hypothetical Arbitrage Example Assume a political market is priced at 40 cents on Kalshi and 45 cents on Polymarket. To hedge, you buy YES on Kalshi and NO on Polymarket. Both directions use actual BUY asks. | Venue | Action | Price | Taker Fee (100 contracts) | Total Cost | | :--- | :--- | :--- | :--- | :--- | | Kalshi | Buy 100 YES | $0.40 | $1.68 | $41.68 | | Polymarket | Buy 100 NO | $0.55 | $0.99000 | $55.99 | | **Total** | **Hedged** | **N/A** | **$2.67** | **$97.67** | In this hypothetical scenario, the gross cost is $97.67 to guarantee a $100.00 payout at resolution. The net profit is $2.33. However, this assumes instant execution across both order books without slippage. If you use resting limit orders to capture maker rebates on Polymarket, execution risk increases. Note that a limit order can be a taker if marketable. ## Are play-money platforms like Manifold useful for real arbitrage? Manifold operates on a play-money system, meaning there is no direct financial arbitrage available. You cannot buy a contract on Manifold and hedge it on Kalshi to lock in a USD profit. However, quantitative traders still monitor these venues closely. Play-money incentives and participation differ from real-money markets. Their prices may supply research signals, but there is no general rule that they lead real-money order books. A sudden spike in a Manifold market can serve as a leading indicator, providing a hypothesis to test against timestamps, liquidity, and out-of-sample results before routing any real-money order. If you are building automated systems, ensure you review our [fake bot checklist](/fake-bot-checklist) to avoid purchasing useless signal scrapers. ## What hidden frictions kill cross-exchange prediction market edge? The most dangerous friction in prediction market arbitrage is resolution risk. Venues can resolve the "same" event differently based on their specific rulebooks. For example, a market on whether a specific piece of legislation will pass might be defined by a Senate vote on one exchange, but require the President's signature on another. If the bill passes the Senate but is vetoed, your perfectly hedged arbitrage trade will lose on both sides. Additionally, capital efficiency is a major hurdle. Crypto venues like Limitless require stablecoins, while Kalshi and ForecastEx require fiat USD. Moving capital between traditional banking rails and blockchain networks incurs bridging fees, wire fees, and time delays. By the time your funds settle, the arbitrage window has usually closed. This is not financial advice, but traders must maintain pre-funded balances on all target exchanges to successfully execute cross-venue strategies. ## Frequently asked questions ### Do maker rebates offset taker fees in cross-exchange arbitrage? Yes, if you provide liquidity on a platform like Polymarket, the maker rebates can subsidize the taker fees paid on the opposing venue. However, resting orders carry execution risk, meaning the market might move away from your price before you are filled. Rebates eligibility is not guaranteed revenue. ### Why do fees peak at 50 cents on Kalshi and Polymarket? The fee formulas use the product of the price and the inverse price `p * (1 - p)`. Mathematically, this quadratic function reaches its maximum when the probability is exactly 50%, making coin-flip markets the most expensive to trade aggressively. ### Can I short a contract on ForecastEx? No, short selling is not permitted. To reduce or exit a position, you must buy the opposing event contract at the same strike, and the broker will net the positions together. ## Sources * [NBER: LIMITS TO ARBITRAGE AND PREDICTION MARKET EFFICIENCY: THE CASE OF PREDICTIT](https://www.nber.org/system/files/working_papers/w35845/w35845.pdf) * [Interactive Brokers: Prediction Markets](https://www.interactivebrokers.com/en/pricing/commissions-events.php) * [Robinhood: event contracts overview](https://robinhood.com/us/en/support/articles/event-contracts-overview/) * [Interactive Brokers: Event Contracts in the Web API](https://www.interactivebrokers.com/campus/ibkr-api-page/event-contracts/) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) * [PredictIt: platform announcements](https://www.predictit.org/platform-announcements) * [PredictIt: deposits and withdrawals update](https://www.predictit.org/stories/detail/102/Product-Update--Deposits-and-Withdrawals-are-Free) --- ### Vetting a Prediction Market Arbitrage Bot Before Connecting Canonical: https://predictionmarketarbitrage.com/blog/vetting-arbitrage-bots-before-connecting-wallet Published: 2026-11-02 Updated: 2026-11-16 Author: PMA Research Summary: Protect your funds from malicious prediction market arbitrage bots. Learn how to audit dependencies, scope permissions, and verify performance. The appeal of automated trading is undeniable, but the infrastructure required to execute it safely is unforgiving. Handing over your credentials to an unverified script is the fastest way to drain your trading capital. Before you connect a wallet to any automated system, you must rigorously audit its code, scope its permissions, and verify its performance claims. ## Key takeaways * Malicious repositories often hide credential-stealing payloads inside obfuscated npm dependencies rather than the main source code. * Legitimate trading automation requires scoped API keys with withdrawal permissions explicitly disabled, never raw private keys. * Backtested performance claims are frequently inflated by ignoring taker fees, which can consume a significant portion of your edge. * Running unverified code in an isolated virtual machine helps mitigate risks, but security VM and burner mitigations are not absolute safety guarantees. ## How do malicious prediction market bots steal private keys? Threat actors have developed sophisticated supply chain attacks targeting cryptocurrency developers and algorithmic traders. Instead of writing obvious malware, attackers hijack legitimate, dormant open-source projects to distribute their payloads. In March 2026, security researchers discovered that a verified GitHub organization belonging to a legitimate Japanese decentralized finance project had been compromised. The attackers used this trusted account to host over twenty malicious repositories disguised as automated trading software [StepSecurity: Malicious Polymarket Bot Hides in Hijacked dev-protocol GitHub Org and Steals Wallet Keys](https://www.stepsecurity.io/blog/malicious-polymarket-bot-hides-in-hijacked-dev-protocol-github-org-and-steals-wallet-keys). Because the organization had a verified badge and a history dating back to 2019, the repositories appeared credible to unsuspecting users [GitHub: Security Alert: This GitHub organization appears to have been compromised](https://github.com/dev-protocol/dev-kit-js/issues/1687). The malicious code was not located in the primary trading logic. The bots actually functioned as advertised, connecting to real market data endpoints and authenticating correctly. However, the repositories included typosquatted npm packages. When a user installed the dependencies, a postinstall hook automatically executed an obfuscated script. This script scanned the local machine for environment files, exfiltrated wallet keys to an external command server, and opened a secure shell backdoor. This is part of a broader trend. Threat intelligence firms have mapped extensive delivery networks operated by state-sponsored actors, identifying dozens of malicious packages embedded across hundreds of repositories targeting various decentralized ecosystems [Panther: Tracking the Contagious Trader Distribution Network](https://panther.com/blog/mapping-the-contagious-trader-delivery-network). Relying on star counts or verified badges is insufficient for security. ## Why should you audit npm dependencies and postinstall scripts? The modern software supply chain is deeply interconnected. A single trading script might rely on hundreds of third-party libraries. Attackers exploit this by publishing packages with names that are one letter off from popular libraries. When you run an installation command, you are granting execution privileges to the authors of every package in that dependency tree. Postinstall scripts are particularly dangerous because they run automatically during the installation phase, before you even execute the main application. To protect yourself, you must inspect the full dependency tree. Look for packages with unusually low download counts or names that mimic established libraries. Review the lockfile to ensure the versions you are installing match the versions that were previously audited. If a repository lacks a lockfile, the dependencies are floating, meaning a benign package could be swapped for a malicious one at any time. ## What is the difference between API keys and private keys? Understanding the distinction between API keys and private keys is critical for securing your funds. A private key is the cryptographic root of your wallet. Anyone who possesses it has total control over your assets and can transfer them to any address. You should never paste a raw private key or seed phrase into a command-line interface or configuration file provided by an unverified third party. API keys, conversely, are generated by the exchange or platform and can be strictly scoped. When generating an API key for an automated strategy, you must enforce the principle of least privilege. Enable trading permissions but explicitly disable withdrawal permissions. While a properly scoped trade-only API key restricts direct withdrawals to an attacker's wallet, if compromised, the attacker might still execute unprofitable trades to drain your balance into their own positions. Always store API keys in secure environment variables or a dedicated secrets manager. Never hardcode them into your source files or commit them to version control. ## How can you safely test a bot in an isolated environment? Even after auditing the code, you should never run unverified software on your primary workstation. Use an isolated virtual machine or a containerized environment to execute the initial setup and testing phases. This isolation helps mitigate the risk of malicious scripts accessing your personal files, browser extensions, or global environment variables. However, remember that security VM and burner wallet mitigations are not absolute safety guarantees. ### Pre-deployment security checklist | Verification Step | Action Required | Rationale | | :--- | :--- | :--- | | Dependency Audit | Inspect the full package tree for typosquatted names. | Prevents execution of hidden malware during installation. | | Network Monitoring | Log all outbound connections during a dry run. | Identifies unauthorized data exfiltration to command servers. | | Permission Scoping | Restrict API keys to trade-only access. | Prevents unauthorized withdrawals if credentials are leaked. | | Environment Isolation | Execute the software in a dedicated virtual machine. | Mitigates risks to your primary workstation from secure shell backdoors. | | Burner Wallet | Fund a separate address with minimal capital. | Caps your maximum potential loss in the event of a compromise. | Run the software without providing any real credentials first. Monitor its network activity to ensure it only communicates with expected endpoints, such as official exchange APIs or public remote procedure call nodes. Any connection to an unrecognized domain warrants immediate investigation. ## Execution Realities: Order Types and Market Data Vendors often market their software using highly optimized backtests that ignore the friction of real-world trading. A strategy that appears profitable on paper will quickly collapse if it fails to account for taker fees, slippage, and execution latency. When evaluating market data for backtests, remember that outcome prices, last trade prices, or midpoints are NOT executable asks. Missing asks mean the contract is unavailable, never zero. Selling YES uses actual bids and collateral/venue permission. Furthermore, a limit order can be a taker if it is marketable; only resting, non-marketable orders provide liquidity. When attempting arbitrage across multiple platforms, remember that Fill-or-Kill (FOK) controls fills on ONE venue, never atomic cross-venue. FOK and Immediate-or-Cancel (IOC) semantics differ significantly, and misunderstanding them can leave you with an unhedged, directional position. ## Verifying Fee Calculations Consider the fee structures of major platforms. Polymarket charges taker fees on most categories, while makers pay no fee and can earn rebates, though rebate eligibility is not guaranteed revenue [Fees - Polymarket Documentation](https://docs.polymarket.com/polymarket-learn/trading/fees). On Kalshi, the general taker fee formula is `round up(M * 0.07 * C * P * (1 - P))`, where M is the multiplier (defaulting to 1 for standard contracts) [Fee Schedule for July 2026 - 7.7.26 Update](https://kalshi.com/docs/kalshi-fee-schedule.pdf). Let us examine a worked numeric example. Suppose you are evaluating an arbitrage opportunity on Kalshi where a YES contract pays $1.00 at resolution. You want to purchase 1,000 contracts at a price of 50 cents ($0.50). * Base cost: 1,000 contracts x $0.50 = $500.00 * Fee calculation: 1 x 0.07 x 1,000 x 0.50 x (1 - 0.50) * Fee calculation: 0.07 x 1,000 x 0.25 = $17.50 The fee is approximately 1.75 cents per contract at 50 cents. Your total cost to enter the position is $517.50. You must explicitly acknowledge that current precision must be read from live fee parameters; do not assume a per-contract 1.75-cent hard cap after rounding or across multipliers. If your backtest ignores this friction, your projected returns will be entirely fictitious. Hypothetical tables are not observed trades; return calculations must include both legs' capital and costs. Furthermore, venues can resolve the "same" event differently based on their specific rulebooks. A strategy that assumes identical resolution criteria across platforms without reading the fine print is fundamentally flawed. Always verify the exact settlement conditions before deploying capital. This is not financial advice. Always conduct your own rigorous due diligence. There is no guaranteed profitability in prediction market arbitrage. ## Frequently asked questions ### Should I trust a repository with a high star count? No. Star counts and forks are easily manipulated using automated accounts. You must evaluate the actual commit history and audit the dependency tree rather than relying on superficial metrics. ### Can a trade-only API key be used to steal my funds? While a properly scoped trade-only API key cannot be used to withdraw funds directly to an attacker's wallet, an attacker could potentially use it to execute trades that benefit their own positions at your expense. ### Why do backtests often overstate profitability? Backtests frequently suffer from look-ahead bias and fail to account for real-world friction. They often ignore taker fees, slippage, the fact that marketable limit orders act as takers, and the reality that you cannot always execute your entire order at the best available historical price. ## Sources * [StepSecurity: malicious Polymarket bot investigation](https://www.stepsecurity.io/blog/malicious-polymarket-bot-hides-in-hijacked-dev-protocol-github-org-and-steals-wallet-keys) * [GitHub: Security Alert: This GitHub organization appears to have been compromised](https://github.com/dev-protocol/dev-kit-js/issues/1687) * [Panther: Tracking the Contagious Trader Distribution Network](https://panther.com/blog/mapping-the-contagious-trader-delivery-network) * [Fees - Polymarket Documentation](https://docs.polymarket.com/polymarket-learn/trading/fees) * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) --- ### Famous Prediction Market Inefficiencies & Arbitrage History Canonical: https://predictionmarketarbitrage.com/blog/historical-prediction-market-inefficiencies Published: 2026-11-02 Updated: 2026-11-16 Author: PMA Research Summary: Explore historical prediction market design, position limits, and pricing frictions, and distinguish directional wins from executable arbitrage evidence. Prediction markets are frequently praised as highly efficient forecasting engines, yet their history is littered with structural anomalies and pricing distortions. For quantitative traders, these historical mispricings are not just academic curiosities; they provide the foundational blueprint for extracting alpha. By examining past market inefficiencies, we can better understand the mechanics of modern arbitrage. ## Key takeaways * Strict position limits and high fees structurally force market prices to deviate from true probabilities, creating persistent negative-sum environments. * Attempted market manipulation typically fails in highly liquid markets, as deep order books allow arbitrageurs to quickly absorb and correct artificial price spikes. * Regulatory fragmentation creates cross-exchange arbitrage opportunities, though traders must account for the risk that venues can resolve the "same" event differently. * Large directional bets often reflect heterogeneous beliefs rather than malicious manipulation, providing liquidity that sharp traders can exploit. ## How did the Iowa Electronic Markets pioneer prediction accuracy? The Iowa Electronic Markets launched in 1988 as a research and teaching tool, later operating under a 1992 regulatory no-action letter. Despite its academic origins, it consistently outperformed traditional polling. The secret to its relative efficiency lay in its unique market design. The exchange imposed a strict $500 investment cap per trader and charged zero fees. Because there were no transaction costs, a YES contract that pays $1.00 at resolution could be traded purely based on its implied probability. The $500 cap limited the financial damage any single irrational or biased trader could inflict on the order book. High-activity algorithmic traders could easily enforce arbitrage bounds without worrying about fee drag. This environment allowed a small group of moderately active, highly informed traders to set efficient prices. The absence of fees meant there was no wedge between the trading price and the actual forecasted probability, a stark contrast to modern commercial platforms where fee structures heavily dictate trading strategies. ## What happened during the 2012 Intrade market manipulation? Intrade was a dominant offshore prediction market in the early 2010s. During the 2012 United States presidential election, the platform experienced one of the most famous attempted market manipulations in prediction market history. In the final two weeks of the race, a single trader accounted for roughly one-third of all money staked on Mitt Romney. On Election Day, this "Romney whale" placed massive buy orders, artificially holding the candidate's odds steady for hours even as prices on rival exchanges moved sharply toward Barack Obama. The distortion persisted primarily because Intrade required traders to post cash covering their worst-case losses upfront. Few arbitrageurs had the capital readily available on the platform to trade against the whale on election night. Once the whale pulled the orders following exit polls, the price immediately collapsed to match other markets. The manipulator lost nearly $7 million. This episode highlights that while manipulation can temporarily distort prices, it ultimately serves as a highly profitable opportunity for capitalized arbitrageurs who are positioned to fade the artificial momentum. ## Why do PredictIt markets often sum over 100 percent? PredictIt is notorious among quantitative traders for markets where the implied probabilities of mutually exclusive outcomes sum to well over 100 percent. This inefficiency is a direct result of the platform's fee structure and regulatory constraints. PredictIt charges a 10 percent fee on trading profits, while historically capping individual investments at $850 per contract (raised to $3,500 in July 2025). (A previous 5 percent withdrawal fee was removed in September 2026, though the profit fee remains a significant friction). These frictions prevent traders from executing risk-free arbitrage when the market overprices a set of candidates. Consider a hypothetical multi-candidate primary election market: | Candidate | YES Price | NO Price | Implied Probability (YES) | | :--- | :--- | :--- | :--- | | Candidate A | $0.45 | $0.58 | 45% | | Candidate B | $0.35 | $0.68 | 35% | | Candidate C | $0.30 | $0.73 | 30% | | **Total** | | | **110%** | In a frictionless market, a trader could buy NO shares on all candidates, guaranteeing a payout because only one candidate can win. However, the 10 percent profit fee eats into the narrow margins of this trade. Furthermore, the historical $850 position limit (now $3,500) restricted the amount of capital a single trader could deploy to correct the mispricing. As a result, the market remains chronically inefficient. Traders must carefully use a [fee calculator](/fee-calculator) to determine if a seemingly obvious arbitrage opportunity is actually profitable after all platform tolls are extracted. ## How did Betfair and Intrade cross-exchange gaps emerge? During the same 2012 election cycle, significant price discrepancies emerged between Intrade and Betfair. Betfair, a London-based betting exchange, hosted a massive presidential market but effectively blocked American users. Intrade, conversely, had a user base where up to 40 percent were United States residents, despite regulatory warnings. This geographic and regulatory fragmentation meant that the two exchanges had different participant demographics, leading to divergent pricing for the exact same real-world event. European traders on Betfair priced the election differently than the American-heavy user base on Intrade. Arbitrageurs who had access to both platforms could theoretically buy the cheaper YES contract on one exchange and sell it on the other. However, cross-exchange arbitrage carries hidden risks. Venues can resolve the "same" event differently based on their specific rulebooks. A candidate conceding might trigger a payout on one platform, while another might wait for official electoral college certification. Understanding these nuances is a critical component of any cross-exchange strategy, a concept detailed further in our [glossary](/glossary). ## What can concentrated election positions teach us? Large directional positions and arbitrage are different strategies. A trader who buys a candidate to win remains exposed to that result, even if the bet is ultimately profitable. A realized windfall is not evidence of a riskless pricing inconsistency, and wallet-level profit estimates should not be treated as audited account statements. Public trading data can show executed transactions and changing prices. Establishing manipulation or its absence requires evidence beyond the fact that a trader was large or won. Other participants may supply opposing capital, but that alone does not prove that prices were unaffected or that a market was efficient. Avoid attributing a motive or a precisely measured price impact without a suitable study and identification strategy. For an arbitrage analysis, the relevant historical evidence is a synchronized set of executable quotes, depth, actual fees, and matching resolution terms. A narrative about a prominent trader cannot replace that record. Keep current-fee examples separate from historical replay, and distinguish research estimates from realized fills. This distinction changes how historical results should be used. A case study can motivate a hypothesis about crowding, capital segmentation, or liquidity. Testing it requires a defined sample and a baseline, including quiet periods and failed opportunities rather than selecting only memorable election nights. Measure how long quoted quantities remained available and whether the same account could actually access both venues. Then include the financing and rebalancing costs required to maintain the positions. A market can correctly anticipate an election winner while particular prices were expensive, and a well-calibrated market can still produce losing trades. Forecast accuracy, trader profitability, and the existence of executable arbitrage are related but distinct claims. Keep those questions separate when deciding what a historical episode actually demonstrates. ## Frequently asked questions ### Why do prediction markets sometimes fail to predict outcomes accurately? Prediction markets aggregate the beliefs of their participants, which can be skewed by strict position limits, high transaction fees, or a lack of diverse participant demographics. When arbitrageurs are restricted from deploying capital to correct mispricings, the market reflects these structural frictions rather than true real-world probabilities. ### How do fees impact arbitrage opportunities? Fees act as a direct tax on trading margins, often making mathematically obvious arbitrage unprofitable in practice. If the combined cost of taker fees, profit fees, and withdrawal fees exceeds the price discrepancy between contracts, the inefficiency will persist because no rational trader will execute the correcting trade. ### Can a single wealthy trader manipulate a prediction market? While a wealthy trader can temporarily distort prices in illiquid markets, deep and highly liquid markets quickly absorb these attempts. Arbitrageurs and market makers will aggressively trade against the artificial price spike, correcting the market and often causing massive financial losses for the manipulator. ## Sources * [arXiv: The Anatomy of a Blockchain Prediction Market: Polymarket in the 2024 U.S. Presidential Election](https://arxiv.org/html/2603.03136v3) * [CNBC: How prediction-market trades can distort election odds](https://www.cnbc.com/2026/10/30/gamble-on-democracy-inside-suspicious-election-contracts-trading-on-prediction-markets-.html) * [arXiv: The Anatomy of Polymarket: Evidence from the 2024 Presidential Election](https://arxiv.org/html/2603.03136v1) --- ### CFTC Regulation of Prediction Markets Explained for Arbitrageurs Canonical: https://predictionmarketarbitrage.com/blog/cftc-regulation-prediction-markets-explained Published: 2026-11-01 Updated: 2026-11-16 Author: PMA Research Summary: Understand documented CFTC milestones, the distinction between Polymarket entities, and the eligibility and contract checks needed before cross-venue trading. Prediction market regulation matters to arbitrage because an apparently attractive spread is useless if a trader cannot lawfully access both contracts or withdraw the proceeds. In the United States, the CFTC oversees registered derivatives exchanges, but a platform's registration is not a blanket endorsement of every product, trading strategy, or user's eligibility. This guide separates documented regulatory milestones from the contract-specific checks that a trader still needs to make. ## Key takeaways * KalshiEX received CFTC designation as a contract market in November 2020; the announcement was published on November 4, following the November 3 designation order. * In January 2022, the CFTC announced a $1.4 million penalty against the operator of Polymarket's then-unregistered event-based binary options market. * The CFTC's registry separately identifies QCX LLC doing business as Polymarket US as a designated contract market. That does not make the international platform interchangeable with the US entity. * Registration, individual market terms, geographic eligibility, identity checks, and current legal restrictions are separate checks, not substitutes for one another. ## What does the CFTC oversee? The Commodity Futures Trading Commission administers the US federal framework for commodity futures and other derivatives within its jurisdiction. A designated contract market, commonly abbreviated DCM, is an exchange designation under that framework. Registration brings obligations for the exchange; it is not a government guarantee that customers will make money or never suffer operational losses. The CFTC's [November 2020 Kalshi announcement](https://www.cftc.gov/PressRoom/PressReleases/8302-20) provides a clear, primary-source record of the exchange's designation. It should not be paraphrased as a declaration that every possible future event contract is approved or that every person worldwide can trade on the exchange. Likewise, the phrase prediction market describes a broad economic activity, not one universal legal classification. Research markets, play-money communities, offshore services, and registered derivatives exchanges can operate under different rules. Begin by identifying the actual legal entity and product rather than assuming similar interfaces have similar regulatory treatment. ## What did the 2022 Polymarket enforcement action establish? On January 3, 2022, the CFTC announced that it had ordered Blockratize, doing business as Polymarket, to pay a $1.4 million civil monetary penalty for offering off-exchange event-based binary options contracts and failing to obtain the required registration. The announced order also required the wind-down of markets that did not comply with the Commodity Exchange Act and applicable regulations. That is a documented historical action, not proof of a new investigation, a current finding against every affiliated entity, or a statement about every country's law. The appropriate primary citation is the [CFTC's enforcement announcement](https://www.cftc.gov/PressRoom/PressReleases/8478-22), together with the order linked from it. For present-day decisions, read the current terms and availability restrictions of the specific service. Do not infer that a crypto wallet removes eligibility requirements, or that publicly readable prices create permission to place trades. Avoid attempts to evade access restrictions; such workarounds can violate terms and expose a trader to legal or account-access problems. ## Why distinguish Polymarket US from the international platform? The CFTC's trading-organization registry lists QCX LLC doing business as Polymarket US as designated, with a July 9, 2025 designation date. The registry and amended designation materials identify the legal entity. They are better evidence of registration than a screenshot, a social-media claim, or a shared brand name. Registration of that US entity does not automatically extend its status to a separately operated international product. Order books, counterparties, settlement arrangements, available contracts, fees, and eligibility can differ. An arbitrage plan must identify exactly which service supplies each leg. This distinction also matters when researching documentation. International Polymarket fee or API documentation should not automatically be applied to a US product. Conversely, a US registration announcement does not establish that a US resident may access every market on the international website. Match every operational assumption to the actual entity and contract being traded. ## Does federal registration settle every legal question? No. The relationship between federal derivatives regulation and state or other applicable laws can be disputed, particularly for event categories that overlap with wagering. Litigation and restrictions can change. This article does not claim a fixed nationwide answer for all sports, political, or other event contracts. A court decision applies to the issues, parties, procedural posture, and relief actually addressed. An interim injunction is not necessarily a final merits decision, and a ruling involving one exchange or category should not be generalized into universal permission. For a live restriction, consult the current order and the platform's current availability notice, not an old headline. The practical implication is conservative: verify lawful access to both legs before allocating capital. If that cannot be established, a mathematically positive spread is not an executable strategy. Registration also does not supply FDIC insurance for the trading position or guarantee settlement under every failure scenario. ## How do regulatory differences affect an arbitrage trade? Consider a purely hypothetical pair of complementary contracts: YES costs $0.45 on one eligible venue and NO costs $0.50 on another. Buying 100 of each costs $95 before fees. A combined $100 settlement would produce a $5 gross margin only if the contracts cover the same event, settle complementarily, and both venues perform as required. Eligibility is the first hurdle. The trader must lawfully access both services and satisfy their identity and account requirements. Contract equivalence is the second: two markets can use different sources, deadlines, recount provisions, cancellation rules, or definitions of an official result. Costs are a third hurdle. Apply each actual product's fees, available depth, funding costs, and withdrawal terms. Our [fee calculator](/fee-calculator) illustrates fee sensitivity, but a formula from one product is not universal. Finally, account for leg risk: an order accepted on one exchange does not guarantee execution on the other. ## What should traders verify before funding accounts? 1. **Legal entity:** Find the operator's exact name and check relevant official registration records. 2. **Personal eligibility:** Review residence, age, identity verification, and product-specific restrictions. Do not assume an exchange is US-only or globally available without checking. 3. **Contract rules:** Save the resolution source, deadline, payout, cancellation, and dispute provisions for both legs. 4. **Operational permissions:** Confirm available order types, position limits, funding assets, and withdrawal conditions for the actual account. 5. **Failure planning:** Decide what happens if trading is suspended, an account is restricted, or a settlement is disputed. Recovery is not guaranteed by a trading spreadsheet. Retain dated copies or references where practical. A later rule change should be distinguishable from the terms that applied when the position was opened. If the rules are ambiguous, seek qualified legal advice rather than using a disclaimer as permission to proceed. ## Frequently asked questions ### Does a DCM designation mean every listed trade is risk-free? No. Registration concerns the regulatory framework, not profitability, perfect contract matching, uninterrupted access, or guaranteed recovery of customer funds. ### Can I infer tax treatment from the CFTC's classification? No. Federal tax character requires its own analysis under applicable tax law. Read our [tax overview](/blog/prediction-market-profits-tax-treatment) and consult a qualified tax adviser. ### Is an accessible API proof that I may trade? No. Public market-data access and authorization to open an account or place orders are different. Check the operator's current terms and applicable restrictions. ## Sources * [CFTC: CFTC Designates KalshiEX LLC as a Contract Market, November 4, 2020](https://www.cftc.gov/PressRoom/PressReleases/8302-20) * [CFTC: Polymarket enforcement announcement, January 3, 2022](https://www.cftc.gov/PressRoom/PressReleases/8478-22) * [CFTC: Designated Contract Markets registry](https://www.cftc.gov/IndustryOversight/IndustryFilings/TradingOrganizations) * [CFTC: Polymarket US amended order of designation](https://www.cftc.gov/media/12806/Polymarket%20US%20Amended%20Order%20of%20Designation/download) --- ### Beyond Two Legs: Prediction Market Arbitrage Strategies Canonical: https://predictionmarketarbitrage.com/blog/intra-market-related-event-arbitrage Published: 2026-10-31 Updated: 2026-11-16 Author: PMA Research Summary: Discover advanced prediction market arbitrage strategies, from multi-outcome rebalancing to related-event logic, and learn where the edge truly lies. Prediction market arbitrage extends far beyond simple cross-exchange price discrepancies. By exploiting intra-market mispricings and logically linked events, quantitative traders can lock in structural edges without taking directional risk. However, these multi-leg strategies require precise execution and a deep understanding of conditional probability. ## Key takeaways * Market rebalancing arbitrage occurs when the sum of mutually exclusive outcome prices deviates significantly from $1.00. * Buying all YES contracts guarantees a payout when the total implied probability falls below 100 percent. * Buying all NO contracts can yield a profit when the market is collectively overpriced and exceeds the embedded market maker spread. * Related-event arbitrage exploits logical inconsistencies between dependent markets, though differing settlement rules can introduce severe risks. ## How do you trade multi-outcome markets summing under 100 percent? In a perfectly efficient prediction market, coherent frictionless probabilities of all mutually exclusive and collectively exhaustive outcomes sum to 1. Executable ask prices need not sum to $1.00. When the sum of executable YES buy asks, including all fees, falls below the matched $1.00 payout, a structural inefficiency known as long market rebalancing arbitrage emerges. Traders can exploit this by purchasing one YES share for every possible outcome. Because exactly one outcome must resolve to YES, the trader is guaranteed to receive a $1.00 payout at resolution. If the combined cost of acquiring all YES tokens is $0.95, the trader locks in a $0.05 gross profit per set, completely independent of real-world events. Academic research into prediction market inefficiencies notes that [arbitrageurs restore market consistency by adjusting positions until probabilities realign with logical constraints](https://drops.dagstuhl.de/storage/00lipics/lipics-vol354-aft2025/LIPIcs.AFT.2025.27/LIPIcs.AFT.2025.27.pdf). These opportunities typically appear during periods of high volatility or immediately following breaking news when liquidity providers pull their orders and the order book becomes thin. ## When should you buy all NO contracts in a prediction market? The inverse scenario occurs when the sum of all YES prices in a mutually exclusive market exceeds $1.00. However, a sum slightly above $1.00 is normal and represents the standard spread charged by market makers. To find a true short market rebalancing arbitrage, the total implied probability must exceed $1.00 by a margin wide enough to cover all trading costs. In this situation, the optimal strategy is to buy NO contracts across the board. If a market features three mutually exclusive outcomes and you buy one NO contract for each, exactly one outcome will resolve to YES (rendering its NO contract worthless) while the other two will resolve to NO (paying $1.00 each). Your total payout is $2.00. All NO requires sum NO BUY ASKS plus fees < n-1, NOT simply sum YES asks > 1. If the combined cost of the three NO BUY ASKS plus fees is less than $2.00, you secure a profit. ## What is related-event arbitrage and conditional coherence? Related-event arbitrage, sometimes called combinatorial arbitrage, involves trading across two or more markets whose outcomes are logically linked. The core principle is that a subset event can never be priced higher than its superset. For example, a candidate winning an election by five or more points is a subset of that candidate winning the election at all. If the market prices the five-point margin at $0.60 and the outright victory at $0.55, the pricing is logically incoherent. Traders can exploit this by buying the underpriced broader event and taking the opposing position on the overpriced narrower event. Another common structure involves deadline ladders. An event occurring "by March" is contained within the same event occurring "by June." The earlier deadline must trade at or below the price of the later deadline. If it does not, a mechanical arbitrage exists. You can review our [glossary](/glossary) for more definitions of conditional probability terms. ## How do you calculate a multi-leg arbitrage trade? Calculating the exact edge in a multi-leg trade requires accounting for every possible state of the world and deducting all applicable fees. Consider two hypothetical cumulative-deadline contracts using the same starting date, rate definition, source, and cancellation rules. Event A is at least one rate cut by March; event B is at least one by June, so A implies B. These are not contracts for separate cuts occurring during March and June. | Market Leg | Position | Execution Price | Potential Payout | | :--- | :--- | :--- | :--- | | Event B (at least one cut by June) | Buy YES | $0.55 | $1.00 | | Event A (at least one cut by March) | Buy NO | $0.40 | $1.00 | Conditional probability multiplication alone does NOT construct a tradable arbitrage; buying A and B is not synthetic AB. Instead, exploit subset A implies B: buy B YES at $0.55 and A NO at $0.40, costing $0.95. The outcomes are: neither happens (payout 1), A only (impossible), B only (payout 2), both happen (payout 1). If the stated implication holds under the actual settlement rules and both legs fill, the minimum contractual payout is $1.00. A positive net margin additionally requires all fees and other costs to remain below $0.05. However, this assumes perfect execution and zero slippage, which is rarely the case in live markets. Always run your numbers through a reliable [fee calculator](/fee-calculator) before executing. ## Why do near-arbitrage opportunities fail in practice? What looks like a risk-free lock on paper often falls apart in production. The most common trap is assuming that two venues will resolve the "same" event identically. Kalshi and Polymarket might list markets on the exact same political race, but their settlement sources and rules regarding candidate replacement or independent runs can differ drastically. If the rules differ, the contracts are not truly equivalent, and the trade is a risky speculation rather than an arbitrage. Fees also destroy marginal edges. Kalshi charges a general taker fee calculated as 0.07 multiplied by the number of contracts multiplied by the price multiplied by (1 minus the price), rounded up to a centicent, though printed fee tables round to cents. Current precision must be read from live fee parameters, and traders should not assume a per-contract 1.75-cent hard cap across all multipliers. Meanwhile, Polymarket charges taker fees across most categories, though makers pay no fee and may be eligible for rebates, which are not guaranteed revenue. Polymarket fees are rounded to five decimal places in USDC. Finally, execution risk is severe. By the time you fill the first leg of a three-leg trade, the prices on the remaining legs may have moved against you. This leaves you holding a naked, directional position you never intended to take. This is not financial advice, but traders must demand a wide margin of safety to absorb slippage and fee drag. ## Frequently asked questions ### What happens if a multi-outcome market does not cover every possibility? If the listed outcomes are not collectively exhaustive, the sum of their prices does not have to equal $1.00. An unlisted "Other" scenario could occur, meaning all your YES contracts would resolve to zero. ### Can I automate related-event arbitrage? Yes, but automated systems must be programmed to read the specific settlement rules of each market, not just the titles. Relying purely on keyword matching will eventually result in buying contracts that settle differently. ### Does a sum of YES asks above $1.00 establish arbitrage? No. Ask prices are purchase prices, not sale proceeds. Buying all NO outcomes requires their actual asks plus fees to be below n minus 1 dollars for n mutually exclusive, collectively exhaustive outcomes. Selling a complete YES set instead requires executable bids, the necessary collateral, and venue permission. ## Sources * [Dagstuhl: Unravelling the Probabilistic Forest: Arbitrage in Prediction Markets](https://drops.dagstuhl.de/storage/00lipics/lipics-vol354-aft2025/LIPIcs.AFT.2025.27/LIPIcs.AFT.2025.27.pdf) * [IMDEA: Unravelling the Probabilistic Forest: Arbitrage in Prediction Markets](https://suarez-tangil.networks.imdea.org/papers/2025aft-arbitrage.pdf) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) --- ### How Breaking News Creates Arbitrage Windows Canonical: https://predictionmarketarbitrage.com/blog/breaking-news-arbitrage-windows Published: 2026-10-30 Updated: 2026-11-16 Author: PMA Research Summary: Discover the mechanics of news-driven mispricings in prediction markets. Learn how latency, liquidity pulls, and fees impact hypothetical arbitrage windows. When a major headline breaks, prediction markets do not instantly synchronize to a single new probability. Instead, a chaotic repricing process unfolds across different venues, creating fleeting hypothetical arbitrage windows that open for seconds to minutes before slamming shut. Understanding the mechanics of these news-driven mispricings separates profitable quantitative strategies from those that merely burn capital on crossing spreads. ## Key takeaways * News shocks cause immediate liquidity pulls as market makers widen spreads to avoid adverse selection. * Decentralized and centralized venues reprice at different speeds due to varying watcher density and institutional participation. * The May 2024 CPI leak demonstrated a 35-minute informational inertia gap between institutional futures and retail prediction markets. * Chasing breaking news requires strict fee accounting, as taker fees on platforms like Kalshi and Polymarket can erase narrow arbitrage margins. ## How do news shocks trigger price discovery? When unexpected information hits the wire, the immediate reaction in any order book is a defensive liquidity pull. Market makers, knowing they are at risk of being picked off by faster participants with better information parsing, cancel their resting orders or drastically widen their bid-ask spreads. This mechanical response ensures that the first aggressive taker orders execute at significantly worse prices than the pre-news top of book. A limit order can be a taker if it is marketable; only resting, non-marketable orders provide liquidity. In prediction markets, a standard winning token typically pays $1.00 at resolution. If a political candidate drops out of a race, the probability of their victory drops to zero. However, the price adjustment is rarely a smooth glide. Academic research on the 2024 U.S. Presidential Election (a non-peer-reviewed preprint) suggests that trading volume spikes after political shocks, driven primarily by incumbent traders rather than new entrants ([Political Shocks and Price Discovery in Prediction Markets: Evidence from the 2024 U.S. Presidential Election](https://arxiv.org/html/2603.03152)). The price response tracks what the news reveals about linked candidates, but the speed of this adjustment depends heavily on the specific venue's microstructure. ## Why do venues reprice at different speeds? The speed at which a market absorbs breaking news depends on watcher density and the composition of its trader base. A market populated by algorithmic traders scanning wire services will reprice in milliseconds. A market dominated by retail participants trading on push notifications may take minutes or even hours to fully reflect the same information. This discrepancy was starkly illustrated during the May 2024 U.S. Bureau of Labor Statistics CPI leak. The data was inadvertently uploaded 30 minutes early. Institutional CME 30-Day Federal Funds futures began adjusting within seconds. In contrast, decentralized retail prediction markets exhibited severe informational inertia. Polymarket contracts showed a near-zero top-of-book response for 35 minutes, only repricing five minutes after the official public release time ([Informational inertia in a decentralized prediction market: Evidence from the May 2024 CPI leak](https://www.sciencedirect.com/science/article/abs/pii/S1544612326009062)). This massive latency gap creates a theoretical arbitrage window, provided a trader can execute across both venues simultaneously. ## What happened during the 2024 political shocks? The 2024 election cycle provided multiple case studies in news-driven mispricings. Three distinct events highlighted how different types of news clear through order books: the first presidential debate, an assassination attempt, and a candidate withdrawal. During the withdrawal event, heavy trading volume hit the market, yet the opposing candidate's YES price remained surprisingly stable. Order-level data from a non-peer-reviewed preprint revealed that taker pressure was highly asymmetric. A small number of large wallets initiated massive exposure reductions, while a larger number of smaller wallets absorbed the flow ([Political Shocks and Price Discovery in Prediction Markets: Evidence from the 2024 U.S. Presidential Election](https://arxiv.org/html/2603.03152v3)). When venues process these shocks asynchronously, cross-exchange spreads open. If Venue A has a YES ask at 40 cents and Venue B has a NO ask at 40 cents, an arbitrageur can theoretically buy YES on Venue A and NO on Venue B, locking in a gross profit of 20 cents per contract. Selling YES uses actual bids and collateral/venue permission; missing asks are unavailable, never 0. Furthermore, this assumes the event resolves identically on both platforms, which is a dangerous assumption as cross-venue opposite labels alone do not guarantee payout. ## How to calculate the true cost of a news arbitrage trade? Gross spreads are an illusion. The true profitability of an arbitrage window must account for execution slippage, capital lockup costs, and exchange fees. Return calculations must include both legs' capital and costs. Consider a hypothetical scenario where a news shock creates a 5-cent spread between Kalshi and Polymarket. You want to execute a 1,000-contract arbitrage trade at a 50-cent price point. * **Kalshi Taker Fee:** The general taker formula is round up(M * 0.07 * C * P * (1 - P)) to a centicent. Assuming a default multiplier M=1, at 50 cents, this is 1 * 0.07 * 1000 * 0.50 * 0.50 = $17.50 (or 1.75 cents per contract). * **Polymarket Taker Fee:** Polymarket fees are calculated as C * feeRate * p * (1 - p), rounded to five decimal places USDC. Assuming a Politics market with a 0.04 fee rate, at 50 cents, this costs 1000 * 0.04 * 0.50 * 0.50 = $10.00000 (or 1.00 cent per contract). ### Arbitrage Profitability Checklist 1. Calculate gross spread: $0.05 per contract. 2. Subtract Kalshi fee: $0.05 - $0.0175 = $0.0325. 3. Subtract Polymarket fee: $0.0325 - $0.0100 = $0.0225. 4. Account for slippage: If average execution price slips by 1 cent on each leg, net profit drops to $0.0025 per contract. Before deploying capital, always run your exact sizing through a [fee calculator](/fee-calculator) to ensure the net yield justifies the execution risk. There is no guaranteed profitability. ## Why is chasing breaking news arbitrage dangerous? Attempting to capture news-driven mispricings is fraught with structural risks. The most immediate danger is stale quotes. When a market maker pulls liquidity, the top-of-book price displayed on your screen may no longer exist by the time your taker order reaches the matching engine. You will either suffer massive slippage or your order will be rejected. FOK (Fill or Kill) controls fills on ONE venue, never atomic cross-venue. Rule ambiguity is another critical failure point. Venues can resolve the "same" event differently based on their specific rulebooks. A geopolitical event might trigger a YES resolution on one platform based on a specific news source, while another platform requires an official government declaration. If you are holding opposing contracts across these venues, you risk losing both legs of the trade. Finally, the infrastructure surrounding these markets is heavily targeted by malicious actors. Numerous fake Polymarket arbitrage bot repos have stolen private keys from unsuspecting traders attempting to automate their news-trading strategies. Always verify your software stack against a reliable [fake bot checklist](/fake-bot-checklist) before granting API access or wallet permissions. Security mitigations are not safety guarantees. This is not financial advice, but basic operational security is mandatory in this space. ## Frequently asked questions ### What causes informational inertia in prediction markets? Informational inertia occurs when a market fails to immediately reprice public information. This is typically caused by low watcher density, a lack of institutional algorithmic participants, or ambiguous event resolution criteria that make traders hesitant to provide liquidity. ### Do market makers pay the same fees as retail traders? No. On platforms like Polymarket, makers are never charged fees and may be eligible for rebates, though rebate eligibility is not guaranteed revenue. This fee structure incentivizes market makers to maintain tight spreads during normal conditions, though they will still widen spreads during volatile news events. ### Can I manually trade news arbitrage windows? Manual execution is extremely difficult. By the time a human reads a headline, logs into multiple exchanges, and routes orders, algorithmic participants have usually closed the profitable spread. Manual traders are more likely to execute against stale quotes and suffer adverse selection. ## Sources * [Arxiv: Political Shocks and Price Discovery in Prediction Markets: Evidence from the 2024 U.S. Presidential Election](https://arxiv.org/html/2603.03152) * [ScienceDirect: Informational inertia in a decentralized prediction market: Evidence from the May 2024 CPI leak](https://www.sciencedirect.com/science/article/abs/pii/S1544612326009062) * [Arxiv: Political Shocks and Price Discovery in Prediction Markets: Evidence from the 2024 U.S. Presidential Election (v3)](https://arxiv.org/html/2603.03152v3) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) --- ### Backtesting Prediction Market Arbitrage with Historical Data Canonical: https://predictionmarketarbitrage.com/blog/backtesting-arbitrage-historical-data Published: 2026-10-30 Updated: 2026-11-16 Author: PMA Research Summary: Learn how to backtest prediction market arbitrage honestly by sourcing historical data, modeling slippage, fees, and avoiding look-ahead bias. Prediction market arbitrage looks like free money on a chart, but historical price data quietly lies to you. Building a backtest that survives contact with live order books requires reconstructing exact market depth, modeling dynamic fees, and ruthlessly eliminating look-ahead bias. If your simulation fills at the mid-price, you are not modeling a strategy; you are modeling a fantasy. ## Key takeaways * Historical price charts cannot prove an order of your size would have filled; you must use order book snapshots to measure available liquidity. * Filling simulated orders at the last traded price or mid-price awards your strategy the spread, artificially inflating backtest returns. * Venues frequently resolve the "same" real-world event differently, meaning cross-exchange arbitrage carries hidden fundamental risk. * Accurate backtesting requires modeling capital lockup periods and dynamic taker fees, which severely impact high-frequency strategies. ## Where can I find historical prediction market data? Sourcing clean data is the first hurdle in quantitative modeling. Unlike traditional equities, prediction markets do not offer standardized, consolidated tape feeds. You must piece together history from venue-specific APIs and third-party archives. Kalshi separates its API into live and historical tiers. To access older records, you must query the historical endpoints. The boundary between live and historical data advances over time, so you must check the current cutoff timestamps using the official endpoints. For example, you can retrieve settled markets older than the cutoff or fetch specific market details using the Kalshi historical endpoints. Kalshi also provides historical candlesticks at 1-minute, 60-minute, and 1440-minute intervals. When using the REST API (e.g., `https://external-api.kalshi.com/trade-api/v2`), note that the `yes_dollars` and `no_dollars` arrays contain dollar prices and fixed-point sizes, sorted ascending with the best bid last. Crucially, only bids are returned; to find the YES ask, you must calculate `1 - best NO bid` [Kalshi: Quick Start Market Data](https://docs.kalshi.com/getting_started/quick_start_market_data). Polymarket data requires a different approach. While the venue offers a price-history endpoint, reconstructing the full order book depth historically is notoriously difficult. Because Polymarket settles on-chain, researchers often rely on public datasets hosted on Hugging Face or GitHub, which can reconstruct executed trades from blockchain logs, but not the complete off-chain resting order book or canceled quotes. Relying solely on trade prints will show you executed transactions but cannot reveal the resting liquidity you need to simulate large fills. ## Why is matching markets across venues so difficult? Cross-venue arbitrage assumes you are trading identical assets. In prediction markets, a standard winning token pays $1.00 at resolution and the losing token pays $0.00, but the conditions triggering that payout vary wildly between exchanges. Venues can resolve the "same" event differently. One exchange might rely on an Associated Press race call to settle an election market, while another waits for congressional certification. Polymarket uses the UMA optimistic oracle for prediction markets (where disputes can take days) and Chainlink TWAP for up-down markets [Polymarket: Resolution](https://docs.polymarket.com/concepts/resolution). If you buy YES on Venue A and NO on Venue B, a disputed outcome could result in both contracts resolving to NO, leaving you with a total loss. When building your historical dataset, you must manually map contract rules, not just market titles. A robust backtest tracks the "leg failure rate," which measures how often one side of an arbitrage pair fails to execute or resolves unexpectedly. If your backtest assumes perfect fungibility between Kalshi and Polymarket contracts without checking the fine print, your results will be dangerously optimistic. ## How do I model order book depth and slippage? The most common mistake in backtesting is using the last traded price to simulate entries and exits. If a market is sitting at 61 cents bid and 65 cents ask, the last trade might have occurred at 63 cents. If your backtest assumes you can buy at 63 cents, you are quietly awarding your strategy the spread on every position. To backtest honestly, you must cross the spread. A buy order crosses to the ask; a sell order crosses to the bid. Furthermore, candlestick data lacks order book depth. A 1-minute candle might show a favorable ask price, but it cannot tell you if there were 10 contracts or 10,000 contracts available at that level. ### Slippage and Latency Checklist * **Never fill at the mid-price:** Always use the bid for selling and the ask for buying. Selling YES uses actual bids and collateral/venue permission. * **Assume conservative slippage:** If depth data is missing, penalize your fill price by several cents based on historical volume. * **Model latency:** Introduce a delay between your signal generation and the simulated fill to account for network routing. * **Calculate fill probability:** If the available size at the touch is smaller than your order, assume a partial fill or a worse average price. ## What is the correct way to calculate fees and capital costs? Arbitrage margins are razor-thin, making fee modeling critical. You cannot use a flat percentage fee across all venues and timeframes. For example, the Kalshi general taker fee is calculated as `round up(M * 0.07 * C * P * (1 - P))`, where M is the multiplier (default 1 for takers), C is the number of contracts, and P is the price. The fee plus position cost is rounded up to a centicent [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf). You must read current precision from live fee parameters rather than assuming a hard cap. Polymarket charges taker fees based on the formula `fee = C * feeRate * p * (1 - p)`, rounded to five decimal places USDC, not cents [Polymarket: Fees](https://docs.polymarket.com/trading/fees). The fee rate varies by category: Crypto is 0.07, Sports/Economics/Culture/Weather/Other is 0.05, Finance/Politics/Mentions/Tech is 0.04, and Geopolitics is 0. Makers are never charged fees. | Metric | Description | Impact on Arbitrage | | :--- | :--- | :--- | | Net edge per contract | Expected profit after all fees and slippage | Determines if a trade is mathematically viable | | Capital-days | Amount of capital locked multiplied by days until resolution | Highlights the opportunity cost of tied-up funds | | Leg failure rate | Percentage of trades where only one side fills | Measures execution risk in cross-venue strategies | Capital-days is a vital metric. A strategy might show a 2 percent return, but if the capital is locked for six months awaiting resolution, the annualized yield is terrible. Always run your strategy through a fee calculator to ensure the net edge justifies the capital lockup. ## How do I prevent survivorship and look-ahead bias? Survivorship bias occurs when your historical dataset only includes markets that resolved cleanly. If you drop voided, delisted, or disputed markets from your universe, you are ignoring the exact scenarios that cause catastrophic losses in live trading. Your backtest must include the ugly cases. Look-ahead bias is equally destructive. This happens when your strategy accidentally uses information that was not available at the time of the trade. For instance, filtering your historical universe to only include markets that eventually reached $1,000,000 in volume means you are using future knowledge to select your trades today. To audit your system, ensure every signal in your backtest is computed strictly from data available at that exact millisecond. This information is provided for educational purposes and is not financial advice. ## Frequently asked questions ### Can I use 1-minute candlesticks for high-frequency arbitrage? No. Cross-venue arbitrage opportunities often open and close within seconds. A 1-minute candle aggregates this activity, hiding the micro-structure of the order book and making it impossible to verify if an order would have filled. ### Why do my backtest profits disappear in live trading? The most likely culprits are filling at the mid-price, ignoring taker fees, and failing to account for order book depth. Live trading forces you to cross the spread and compete with other participants for limited liquidity. ### How do I get historical data for markets that are no longer active? You must use dedicated historical API endpoints. For example, Kalshi archives settled markets, which can be accessed by checking the historical cutoff timestamps and querying the appropriate historical routes. ## Sources * [Kalshi: Quick Start Market Data](https://docs.kalshi.com/getting_started/quick_start_market_data) * [Polymarket: Resolution](https://docs.polymarket.com/concepts/resolution) * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) --- ### Prediction Market Arbitrage: Leg Risk and Latency Canonical: https://predictionmarketarbitrage.com/blog/leg-risk-partial-fills-latency Published: 2026-10-29 Updated: 2026-11-16 Author: PMA Research Summary: Prediction market arbitrage is rarely risk-free. Discover how partial fills, latency, and leg risk expose traders to severe execution costs. Arbitrage in prediction markets is frequently pitched as a guaranteed return by influencers and automated system vendors. In reality, the milliseconds between routing your first and second order expose you to severe execution risks that can instantly turn a theoretical profit into a directional loss. ## Key takeaways * Cross-venue orders are not atomic, meaning the first leg can fill while the second leg moves, rejects, or partially fills. Fill-or-Kill (FOK) controls fills on one venue only, never atomically across venues. * Retail trading stacks using standard REST APIs (which can cache and lag) face latency disadvantages against professional setups using authenticated WebSockets. * Partial fills on thin order books leave traders with unhedged directional exposure that requires immediate mitigation. * Settlement delays, withdrawal frictions, and precise taker fees can lock up capital and drastically reduce the annualized return of a seemingly profitable spread. ## How does order sequencing create leg risk? Prediction markets operate on central limit order books where multi-leg arbitrage fills sequentially. There is no single mechanism to buy a cross-venue bundle simultaneously. You send the first leg, it fills, and that fill becomes public information that moves the market before your second leg arrives. Consider a hypothetical scenario where you attempt to buy a YES contract at 30 cents on one venue and a NO contract at 30 cents on another. A standard binary YES contract usually pays $1.00 at resolution (though rare Unknown/50-50 outcomes may redeem at $0.50). A combined 60-cent entry theoretically guarantees a 40-cent gross profit. However, the moment your YES order fills, the market reacts. By the time your NO order reaches the second venue, the best ask has shifted to 78 cents. Your total cost is now 108 cents for a $1.00 payout, resulting in an 8-cent loss. The pre-execution guarantee becomes a post-execution directional bet the instant the first leg prints. ## What happens during a partial fill on thin books? Liquidity depth is often an illusion in fast-moving markets. The displayed best price may only support a small size, meaning larger orders will consume worse price levels or fail to fill entirely. A limit order can be a taker if it is marketable. Missing asks are unavailable, never zero, and both directions of an arbitrage require buying actual asks. Suppose you spot a hypothetical opportunity and attempt to execute 100 contracts on both sides. You successfully purchase 100 YES contracts at 46 cents. However, the opposing book is thin, and you only secure 70 NO contracts at 48 cents before the price moves away. * Matched bundle: 70 contracts fully hedged. * Unmatched exposure: 30 YES contracts carrying full directional market risk. If the market moves against your unhedged YES position, the losses on those 30 contracts can easily wipe out the gross edge of the 70 matched pairs. You must immediately decide whether to close the unhedged YES contracts at a loss or hold them and accept the directional risk. ## How do latency and API rate limits impact execution? Latency is the silent killer of arbitrage strategies. The total time from spotting an opportunity to confirmation breaks down into detection, computation, submission, and confirmation. Retail traders using browser automation or standard REST API setups face a massive disadvantage. For example, Kalshi's public REST market data requires no authentication but responses are cached and can lag by seconds, whereas WebSocket connections require authentication and provide faster updates. Professional bots utilizing optimized connections will consume the available liquidity long before a retail order arrives. Furthermore, API rate limits and connectivity outages can leave you stranded. If you submit your first leg and the second venue's API rate-limits your hedge order, you are left holding a naked position. Automation introduces additional failure modes, including software bugs and clock drift. ## Why do settlement timing and withdrawal frictions matter? Even if you successfully execute both legs, your capital is locked until the market resolves. Venues can resolve the "same" event differently based on their specific rulebooks, not just the market title. For instance, Polymarket uses the UMA optimistic oracle where disputes can take days, while other platforms might use different data sources. Additionally, you must account for the specific fee structures of each platform. The Kalshi general taker fee formula is `round up(M * 0.07 * C * P * (1 - P))`, where the fee plus position cost is rounded up to a centicent ([Fee Schedule for July 2026 - 7.7.26 Update](https://kalshi.com/docs/kalshi-fee-schedule.pdf)). For example, the general fee for 100 contracts at 51 cents is an unrounded 1.7493, though printed tables may display about $1.75. Meanwhile, Polymarket charges taker fees based on the category (e.g., 0.07 for Crypto, 0.05 for Sports/Combos, 0 for Geopolitics) using the formula `fee = C * feeRate * p * (1 - p)`, rounded to five decimal places in USDC ([Fees](https://docs.polymarket.com/trading/fees)). For instance, 100 contracts at 0.46 with a 0.05 rate incurs a 1.24200 USDC fee, not exactly 1.24. Makers are never charged fees, and while they may be eligible for rebates, these are not guaranteed revenue. ## How to hedge a broken leg and manage venue outages? When a leg breaks, you need a predefined failure policy. You cannot afford to hesitate when holding unhedged risk. ### Execution Checklist * Verify contract equivalence and settlement rules before trading. * Check available liquidity depth on both venues, not just top-of-book prices. * Confirm API connectivity and current rate limit status. * Calculate net spread including all taker fees (matching specific venue formulas) and network costs. * Define a maximum acceptable loss for closing an unhedged leg. ### Broken Leg Scenario Table | Failure Mode | Immediate Action | Secondary Risk | | :--- | :--- | :--- | | Partial fill on Leg 2 | Attempt to fill remainder at worse price up to loss cap. | Whipsaw pricing during retry. | | Leg 2 API timeout | Poll order status endpoint to confirm if fill occurred. | Double-execution if retried blindly. | | Venue 2 outage | Halt new trades; attempt to close Leg 1 on Venue 1. | Leg 1 liquidity dries up. | | Leg 2 price spikes | Cancel remaining Leg 2 order; close Leg 1 immediately. | Taking a guaranteed loss on Leg 1. | This is not financial advice. Always test your execution logic in a simulated environment before committing real capital. ## Frequently asked questions ### Can settlement rules break a prediction market arbitrage? Yes. Venues can resolve the "same" event differently if they use different data sources, deadlines, or dispute processes. Rules define resolution, not the title. A trade that looks economically matched can fail to offset if one venue voids the market while the other resolves it normally. ### Why do public scanners show arbitrage opportunities that do not exist? Public scanners often look only at top-of-book prices without accounting for liquidity depth, precise taker fees, or network costs. They may also incorrectly assume missing asks are zero (they are unavailable) or fail to recognize that both directions require buying actual asks. ### How should I handle a partial fill? You must treat the unmatched portion as a naked directional bet. Your automated system or manual process should immediately attempt to close the unhedged contracts or complete the hedge at a worse price, strictly adhering to a predefined maximum loss limit. ## Sources * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) --- ### The Favorite-Longshot Bias in Prediction Markets, Explained Canonical: https://predictionmarketarbitrage.com/blog/favorite-longshot-bias-prediction-markets Published: 2026-10-28 Updated: 2026-11-16 Author: PMA Research Summary: Explore evidence for favorite-longshot bias, its limits across samples and categories, and how fees and measurement choices affect apparent trading edges. Prediction markets offer a brutal lesson in human psychology: traders consistently overpay for low-probability miracles while underpricing near-certainties. This phenomenon, known as the favorite-longshot bias, drains capital from casual participants and creates structural opportunities for disciplined liquidity providers. Understanding the mechanics behind this pricing anomaly is essential for anyone treating event contracts as a serious asset class rather than a lottery ticket. All win rates in the numerical example below are assumptions chosen to illustrate the return formula, not measured venue-wide frequencies. ## Key takeaways * The favorite-longshot bias is a persistent behavioral anomaly where cheap prediction market contracts win far less often than their implied probabilities suggest. * Some evidence, particularly from horse racing, favors probability misperceptions over a simple risk-love explanation; prediction-market mechanisms can differ. * Trading fees and capital lockup disproportionately punish longshot buyers, reducing expected returns without making any individual outcome certain. * Market makers can exploit this bias by systematically fading longshots, though position limits and measurement errors require careful risk management. ## What causes the favorite-longshot bias in prediction markets? The favorite-longshot bias is not a new discovery. It has been documented for decades in horse racing and sports betting, but its presence in modern prediction markets is striking. When a YES contract trades at 5 cents, the market implies a 5% probability of the event occurring. Whether those contracts win less than 5% of the time is an empirical question that depends on the sample, quote timing, weighting, and category. Researchers have long debated the root cause. Is it that traders simply love risk and are willing to pay a premium for a lottery-like payoff? Or do they genuinely misunderstand probabilities? In horse-racing data, a peer-reviewed study on the favorite-longshot bias argues that misperceptions of probability, consistent with prospect theory, better explain the observed pattern than a simple rational risk-love account in that setting [Journal of Political Economy: Explaining the Favorite–Long Shot Bias: Is it Risk-Love or Misperceptions?](https://www.journals.uchicago.edu/doi/10.1086/655844). That result does not establish a single explanation for every prediction market. Risk preferences, market design, fees, and measurement choices can also matter. ## How do fees and capital lockup impact cheap longshots? The mathematical disadvantage of buying longshots is compounded by market friction. When you buy a cheap contract, you are not just fighting bad odds; you are fighting the fee structure and the opportunity cost of capital. Consider the fee mechanics. On Kalshi, the general taker fee formula is `round up(M * 0.07 * C * P * (1 - P))`, where M is a multiplier (defaulting to 1 for takers). While printed fee tables display a peak of about $1.75 for 100 contracts at 50 cents, the actual fee plus position cost is rounded up to a centicent. Traders must explicitly read current precision from live fee parameters and should not assume a strict 1.75-cent hard cap across all multipliers. On Polymarket, taker fees apply to most categories (e.g., 0.07 for Crypto, 0.05 for Sports, 0.04 for Politics) and are calculated as `C * feeRate * p * (1 - p)`. Crucially, Polymarket fees are rounded to five decimal places USDC, not cents. For a contract priced at 5 cents, the fee might seem small in absolute terms, but it represents a massive percentage drag on the expected value of a low-probability bet. If you want to calculate the exact drag on your portfolio, you can use our [fee calculator](/fee-calculator). Furthermore, buying a longshot locks up your capital until resolution. If you tie up funds for six months on a 2-cent contract that has a true probability of 0.5%, you are suffering a severe opportunity cost. ## Are longshots always overpriced across all event types? While the bias is prevalent, it is not uniform across all categories or market regimes. The structure of the event and the macroeconomic environment can significantly alter the pricing dynamics. Recent analysis of Polymarket data in an arXiv preprint reveals that the bias depends heavily on how contracts are grouped. When evaluating individual child markets, purchases below 10 cents lose heavily. However, the pattern varies by category; while crypto and politics show strong longshot losses, sports markets sometimes exhibit different dynamics [arXiv: The Favorite–Longshot Bias in Prediction Markets: Evidence from Polymarket](https://arxiv.org/html/2609.12878v1). Similarly, macroeconomic cycles influence trader behavior. An SSRN working paper on Kalshi's CPI prediction markets found that the favorite-longshot bias actually reversed depending on the Federal Reserve's rate cycle. During the hiking cycle, longshots were overpriced, but during the pause and cutting cycle, longshots became underpriced [SSRN: The Favorite-Longshot Bias in CPI Prediction Markets: Evidence from the Full Sample and Across Fed Rate Cycles](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7093759). This highlights the danger of assuming a static bias across all environments. ## How can market makers and arbitrageurs exploit the bias? For quantitative traders and market makers, the favorite-longshot bias represents a structural edge, though it is distinct from riskless arbitrage. Because makers are never charged fees on Polymarket, they are positioned to take the other side of retail's lottery-ticket demand. However, rebate eligibility is not guaranteed revenue; fills on 0% fee markets (like Geopolitics) earn no maker rebate. Furthermore, there is no universal blanket maker-free claim across all venues. Kalshi charges maker fees on certain series using the formula `round up(M * 0.0175 * C * P * (1 - P))`, where a default maker M=0 applies, but exceptions exist (e.g., M=1 yields a quarter of the taker rate, and certain combos use M=2). By systematically selling YES contracts at 5 cents (or buying NO contracts at 95 cents), makers attempt to capture the behavioral premium. However, this strategy is not without risk. Position limits can severely restrict the ability of arbitrageurs to correct mispricings. An NBER working paper on PredictIt demonstrated that strict position limits prevented sophisticated traders from fully arbitraging away the massive overpricing of longshots, leading to highly inefficient markets [NBER: Limits to Arbitrage and Prediction Market Efficiency: the Case of PredictIt](https://www.nber.org/system/files/working_papers/w35845/w35845.pdf). Additionally, adverse selection is a constant threat. Market makers must ensure they are not providing liquidity to insiders with asymmetric information. Another NBER working paper on Kalshi highlighted that behavioral surplus is often concentrated in single-name markets, requiring makers to be highly selective [NBER: Adverse Selection in Prediction Markets: Evidence from Kalshi](https://conference.nber.org/conf_papers/f243677.pdf). Always consult our [fake bot checklist](/fake-bot-checklist) before deploying automated strategies to fade longshots. ### Illustrative Calibration Table The following hypothetical table demonstrates how the favorite-longshot bias manifests in a typical prediction market environment. | Implied Probability (Price) | Assumed Win Rate (Illustrative) | Expected Return (Pre-Fee) | | :--- | :--- | :--- | | 2¢ (2%) | 0.5% | -75.0% | | 5¢ (5%) | 2.5% | -50.0% | | 10¢ (10%) | 8.0% | -20.0% | | 50¢ (50%) | 50.5% | +1.0% | | 90¢ (90%) | 92.0% | +2.2% | *Note: This table is strictly hypothetical, not a historical reconstruction, and is for illustrative purposes only. It does not constitute financial advice. Return calculations in practice must include both legs' capital and costs. A YES contract pays $1.00 at resolution.* ## Does measurement error exaggerate the longshot tax? Before aggressively shorting every cheap contract, quants must account for data artifacts. The apparent magnitude of the favorite-longshot bias can sometimes be an illusion created by poor measurement techniques. For example, using the last traded price before a market closes can severely distort calibration metrics. In sports markets, the last trade often occurs when the game is already decided, meaning the price reflects certainty rather than a probabilistic forecast. When researchers corrected for this by measuring prices strictly before the outcome was known, the massive bias in certain Kalshi markets practically disappeared, illustrating how quote timing can change an estimated bias [SSRN: Why the Last Trade Is the Wrong Price: Measuring Favorite-Longshot Bias on Kalshi](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7658058). ## Frequently asked questions ### What is the favorite-longshot bias? It is a behavioral pricing anomaly where low-probability events (longshots) are priced higher than their actual frequency of occurrence, while high-probability events (favorites) are slightly underpriced. This results in systematic losses for traders who frequently buy cheap contracts. ### Why do people buy overpriced longshots? Academic research suggests it is primarily due to the misperception of probabilities. Human beings struggle to accurately evaluate very small numbers, often treating a 1% chance and a 5% chance as functionally identical, leading them to overpay for the 5% contract. ### Can I guarantee a profit by shorting longshots? No. There is no guaranteed profitability. While the expected value may be mathematically in your favor, you face capital lockup, trading fees, and the risk of adverse selection. Furthermore, venues can resolve the "same" event differently, introducing settlement risk to any strategy. ## Sources * [Journal of Political Economy: Explaining the Favorite–Long Shot Bias: Is it Risk-Love or Misperceptions?](https://www.journals.uchicago.edu/doi/10.1086/655844) * [arXiv: The Favorite–Longshot Bias in Prediction Markets: Evidence from Polymarket](https://arxiv.org/html/2609.12878v1) * [SSRN: The Favorite-Longshot Bias in CPI Prediction Markets: Evidence from the Full Sample and Across Fed Rate Cycles](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7093759) * [NBER: Limits to Arbitrage and Prediction Market Efficiency: the Case of PredictIt](https://www.nber.org/system/files/working_papers/w35845/w35845.pdf) * [NBER: Adverse Selection in Prediction Markets: Evidence from Kalshi](https://conference.nber.org/conf_papers/f243677.pdf) * [SSRN: Why the Last Trade Is the Wrong Price: Measuring Favorite-Longshot Bias on Kalshi](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7658058) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) --- ### Implied Probability, Overround and Dutch Books in Arbitrage Canonical: https://predictionmarketarbitrage.com/blog/implied-probability-overround-dutch-book-math Published: 2026-10-26 Updated: 2026-11-16 Author: PMA Research Summary: Master the math of prediction market arbitrage. Learn to convert prices to implied probabilities, calculate overround, and compute fee-adjusted break-evens. Prediction market arbitrage relies on a strict mathematical reality: probabilities of mutually exclusive and collectively exhaustive outcomes sum to 100 percent. When prices diverge from this fundamental law across different venues, traders can theoretically lock in a risk-free return, provided they correctly account for trading fees and settlement risks. This guide breaks down the exact formulas required to convert market prices into implied probabilities, identify structural inefficiencies, and calculate your true break-even point. ## Key takeaways * A standard binary prediction market contract priced at 50 cents implies a 50 percent probability of the event occurring, as a winning YES contract usually pays $1.00 at resolution, though partial payouts and ranges exist [Prediction Markets | CFTC](https://www.cftc.gov/LearnandProtect/PredictionMarkets). * The overround, or vig, is the amount by which the sum of all implied probabilities in a market exceeds 100 percent, representing the exchange's or market maker's theoretical margin. * De Finetti's Dutch book theorem proves that if a set of betting quotients violates basic probability axioms, an arbitrageur can construct a series of trades that guarantees a net loss to one side [Stanford Encyclopedia of Philosophy: Dutch Book Arguments](https://plato.stanford.edu/entries/dutch-book). * Cross-venue arbitrage requires calculating the fee-adjusted break-even, as platforms like Kalshi and Polymarket charge quadratic fees that peak when contracts are priced near 50 cents. ## How do you convert prediction market prices to implied probabilities? In a standard binary prediction market, a YES contract typically pays $1.00 if the event occurs and $0.00 if it does not. Because the payout is usually fixed at one dollar, the price of the contract directly represents the market's assessed probability of the outcome. If a YES contract trades at 60 cents, the implied probability of the event happening is 60 percent. This direct translation makes prediction markets highly intuitive compared to traditional sports betting odds. You do not need complex fractional or American odds converters. The price is an implied probability, not proof of the true probability. However, this simplicity can mask underlying market mechanics. When you buy a contract, you are not just taking a position on an outcome; you are trading against the aggregate beliefs of all other market participants. Academic research highlights that prediction market prices are heavily influenced by fee structures, which can cause contract prices for low-probability outcomes to trade below their true probability, creating a unique form of favorite-longshot bias [Taylor & Francis Online: Full article: On prices and returns in commercial prediction markets](https://www.tandfonline.com/doi/abs/10.1080/14697688.2023.2257756). Understanding this bias is critical for anyone attempting to build systematic trading strategies, as the raw price only tells part of the story. Note that favorite-longshot bias is a statistical pricing phenomenon, not a guarantee of riskless arbitrage. ## What is the overround and how does it affect fair value? In a perfectly efficient, frictionless market, the implied probabilities of all mutually exclusive and exhaustive outcomes would sum to exactly 1.0, or 100 percent. In reality, they almost always sum to more than 100 percent. This excess is known as the overround, booksum, or vig. If a market asks 52 cents for YES and 51 cents for NO, the total booksum is 1.03, or 103 percent. That extra 3 percent is the overround. It represents the friction in the market, often driven by market makers widening their bid-ask spreads to protect against adverse selection from informed traders. To find the "fair" probability of an event, quants typically remove the overround by dividing each implied probability by the total booksum. In our example, the proportionally normalized probability for YES would be 0.52 / 1.03, which equals roughly 50.48 percent. This proportional normalization is a comparison convention, not proof of fair value or an executable quote. Actual buy and sell prices, fees, and depth must still be used for trading. ## What is de Finetti's Dutch Book theorem? The mathematical foundation of arbitrage rests on coherence. In probability theory, a set of beliefs or prices is considered coherent if it does not allow for a guaranteed risk-free loss. This concept was formalized by Bruno de Finetti. According to the Stanford Encyclopedia of Philosophy, de Finetti's Dutch book theorem states that given a set of betting quotients that fails to satisfy the probability axioms, there is a set of bets with those quotients that guarantees a net loss to one side [Stanford Encyclopedia of Philosophy: Dutch Book Arguments](https://plato.stanford.edu/entries/dutch-book). For mutually exclusive, collectively exhaustive outcomes, frictionless two-way betting prices that violate the probability axioms permit a Dutch book. Real order books instead have separate bids, asks, fees, and execution constraints. An arbitrageur can theoretically exploit this incoherence by buying a proportional stake in every possible outcome. Because one outcome must occur, the arbitrageur is guaranteed a payout (usually $1.00). If the total cost to acquire that guaranteed dollar is less than $1.00, the arbitrageur has executed a successful Dutch book. This theorem is the theoretical bedrock of all prediction market arbitrage strategies. ## How do you calculate a Dutch book across a 3-outcome market? Arbitrage is not limited to binary YES or NO markets. It frequently appears in multi-outcome markets, such as political primaries or sporting tournaments. Let us walk through a strictly hypothetical 3-outcome market across two different venues to illustrate how a Dutch book is constructed. Suppose we have three candidates (A, B, and C) and we find the following hypothetical best available ask prices across Venue X and Venue Y: | Outcome | Venue | Hypothetical Ask Price | Implied Probability | | :--- | :--- | :--- | :--- | | Candidate A | Venue X | $0.45 | 45% | | Candidate B | Venue Y | $0.30 | 30% | | Candidate C | Venue X | $0.20 | 20% | If we buy one contract for each candidate at these hypothetical prices, our total outlay is $0.45 + $0.30 + $0.20 = $0.95. Assuming standard rules where exactly one candidate must win and pays $1.00, one of these contracts will resolve to $1.00. We have spent 95 cents to guarantee a return of one dollar, locking in a 5-cent gross profit per set of contracts. However, this is strictly a gross calculation. In the real world, you must account for trading fees, withdrawal costs, and the risk that venues can resolve the "same" event differently based on their specific rulebooks. For example, Fill-or-Kill (FOK) orders control fills on ONE venue, but can never provide atomic execution across multiple venues simultaneously. Always read the fine print before deploying capital. For more definitions on market mechanics, consult our [glossary](/glossary). ## How do you compute the fee-adjusted break-even for cross-venue arbitrage? The most common mistake novice arbitrageurs make is ignoring fees. A gross spread of 3 cents might look like a profitable Dutch book, but quadratic fee structures can instantly turn a winning trade into a loss. Consider the fee models of the major platforms. The Kalshi general taker fee is calculated as `round up(M * 0.07 * C * P * (1 - P))`, where M is a multiplier (default 1) [Fee Schedule for July 2026 - 7.7.26 Update](https://kalshi.com/docs/kalshi-fee-schedule.pdf). Meanwhile, Polymarket charges taker fees on most categories while makers pay no fee and can earn rebates [Fees](https://docs.polymarket.com/trading/fees). To compute the fee-adjusted break-even, you must calculate the exact fee for each leg of the trade. Explicitly acknowledge that current precision must be read from live fee parameters; do not assume a per-contract 1.75-cent hard cap after rounding or across multipliers. Let us assume a hypothetical scenario where you find YES on Kalshi at 51 cents and NO on Polymarket at 46 cents for a 100-contract lot. Gross cost = $51.00 + $46.00 = $97.00. Gross profit = $3.00. Now, calculate the fees BEFORE venue rounding: Kalshi fee = `100 * 0.07 * 0.51 * (1 - 0.51)` = 1.7493 [Fee Schedule for July 2026 - 7.7.26 Update](https://kalshi.com/docs/kalshi-fee-schedule.pdf). Polymarket fee (assuming a standard 0.05 rate for this category) = `100 * 0.05 * 0.46 * (1 - 0.46)` = 1.24200 USDC [Fees](https://docs.polymarket.com/trading/fees). Total fees before rounding = 1.7493 + 1.24200 = 2.9913. Net profit before venue rounding = $3.00 - 2.9913 = $0.0087. Total capital deployed = $97.00 + 2.9913 = $99.9913. What looked like a 3-cent gross edge per pair leaves only $0.0087 across all 100 pairs before Kalshi venue rounding, slippage, funding costs, or other frictions. As noted in academic literature, fees are a significant barrier to arbitrage, especially when probabilities are fragmented across many outcomes [NBER: LIMITS TO ARBITRAGE AND PREDICTION MARKET EFFICIENCY](https://www.nber.org/system/files/working_papers/w35845/w35845.pdf). Furthermore, be extremely cautious of automated solutions claiming to execute these trades for you. Fake Polymarket arbitrage bot repos have stolen private keys from unsuspecting users. Always verify code manually and review our [fake bot checklist](/fake-bot-checklist) before connecting any wallets. This is not financial advice, but basic operational security is mandatory in this space. ## Frequently asked questions ### How do you calculate implied probability from a contract price? In a market where a winning contract pays exactly $1.00, the implied probability is simply the price of the contract. A contract trading at 75 cents implies a 75 percent probability of the event occurring. Note that not all contracts universally pay $1.00. ### What is the difference between overround and margin? The overround is the amount by which the sum of all implied probabilities in a market exceeds 100 percent. The margin is the actual percentage of total betting volume the exchange or market maker expects to keep as profit, derived from that overround. ### Why do prediction market prices sometimes fail to sum to exactly $1.00? Prices often fail to sum to $1.00 due to trading fees, bid-ask spreads, and capital constraints that prevent arbitrageurs from instantly closing the gap. If the gross spread is smaller than the combined fees of trading both sides, the mathematical inefficiency will persist. ## Sources * [Prediction Markets | CFTC](https://www.cftc.gov/LearnandProtect/PredictionMarkets) * [Stanford Encyclopedia of Philosophy: Dutch Book Arguments](https://plato.stanford.edu/entries/dutch-book) * [Taylor & Francis Online: Full article: On prices and returns in commercial prediction markets](https://www.tandfonline.com/doi/abs/10.1080/14697688.2023.2257756) * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) * [NBER: LIMITS TO ARBITRAGE AND PREDICTION MARKET EFFICIENCY](https://www.nber.org/system/files/working_papers/w35845/w35845.pdf) --- ### Election Arbitrage Reconstruction: A Hypothetical Fee-Adjusted Case Study Canonical: https://predictionmarketarbitrage.com/blog/reconstructed-election-night-arbitrage-case-study Published: 2026-10-25 Updated: 2026-11-16 Author: PMA Research Summary: A hypothetical election-market trade illustrates fee arithmetic, capital lockup, and settlement risk. It is not a record of observed 2024 fills or realized profits. During the final weeks of the 2024 United States presidential race, prediction markets experienced unprecedented volume and glaring pricing inefficiencies. A hypothetical cross-venue trade can illustrate the difference between a directional election bet and a matched position, without claiming any particular trader achieved the modeled fills. This reconstructed case study breaks down the mechanics, costs, and risks of a hypothetical cross-exchange arbitrage trade to illustrate the realities of market neutrality. This is an educational reconstruction, not a record of observed fills or realized profits. The worked fee model uses the stated current assumptions, not the historical fee schedule of election night 2024. ## Key takeaways * Cross-venue election prices can differ, but a historical opportunity requires timestamped executable quotes, fees, and matched settlement terms, not an election narrative alone. * Execution costs, including Kalshi taker fees and Polymarket's fee structure, severely compress theoretical gross margins. * Capital lockup until event resolution introduces significant opportunity costs that must be modeled against risk-free rates. * Contract resolution criteria often differ across platforms, creating hidden fundamental risks for seemingly identical markets. ## Why did 2024 election prediction markets diverge? Election markets can draw different participants and face different trading frictions. That can motivate a cross-venue comparison, but this article does not establish a historical executable spread from press coverage. A genuine replay would require synchronized quotes, available depth, applicable historical fees, and the same settlement conditions. These discrepancies arise from structural differences between platforms. Kalshi operates as a regulated United States exchange whose participant eligibility depends on current account and geographic restrictions. Polymarket operates on a blockchain infrastructure and historically catered to an international user base. This geographic and regulatory segmentation prevents seamless capital flow. When a sudden news shock occurs, participants on one platform might react faster or with more capital than those on another. Academic analysis confirms that substantial price discrepancies persist across major prediction platforms, with leadership in price discovery varying by event type and trading volume, according to [SSRN: Price Discovery and Trading in Modern Prediction Markets](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5331995) (working paper). Changes in participation can alter depth and price impact, but that mechanism should be tested rather than assumed from election trading volume. A [preprint studying Polymarket during the 2024 election](https://arxiv.org/html/2603.03136v2) examines market development. Any measured convergence time depends on its sample and methodology; it is not a universal window available to an arbitrage trader. However, brief windows of opportunity remained for automated systems capable of executing simultaneous orders. ## How do you calculate fees and size a cross-venue trade? Executing a cross-venue arbitrage requires precise fee modeling. A theoretical price gap of three cents might look profitable until transaction costs are applied. On Kalshi, the general taker fee formula is `round up(M x 0.07 x C x P x (1-P))`, where M is the multiplier (default 1), C is the number of contracts, and P is the price. This creates a fee of approximately 1.75 cents per contract when the price is at 50 cents, as documented in the [Kalshi Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf). Polymarket charges taker fees on most categories, while makers pay no fee and can earn rebates. The fee formula is `fee = C × feeRate × p × (1 - p)`, rounded to five decimal places in USDC. For politics, the taker fee rate is 0.04. For a taker executing an immediate arbitrage, these combined fees eat directly into the gross spread. Traders must use a precise fee calculator to determine the net expected value before routing orders. Sizing the trade depends on the available liquidity at the top of the order book on both venues. If Kalshi shows 5,000 contracts available at 45 cents and Polymarket shows 2,000 contracts available at 50 cents for the opposing outcome, the maximum executable size without slippage is 2,000 contracts. ## What are the hidden risks of prediction market arbitrage? The most dangerous assumption in cross-venue trading is that identical titles mean identical contracts. Venues can resolve the "same" event differently based on their specific rulebooks. For example, one platform might resolve a presidential winner market based on the Associated Press call, while another might wait for the official electoral college certification in January. If a recount or legal challenge delays the official certification, a trader could face a scenario where one leg of the arbitrage pays out immediately while the other remains locked for months. Security risks also plague the decentralized finance ecosystem. The rise of automated trading has led to a proliferation of malicious software. Fake Polymarket arbitrage bot repos have stolen private keys from unsuspecting developers who downloaded unverified code from GitHub. Traders must build their execution infrastructure from scratch or rigorously audit open-source tools using a strict fake bot checklist. ## How does capital lockup impact arbitrage profitability? A standard YES contract typically pays $1.00 if its specified event occurs and zero otherwise; the matched NO contract pays for the opposite outcome. Unlike traditional equity arbitrage where positions can be unwound for a profit once prices converge, prediction market arbitrage often requires holding the contracts until the event actually occurs. This creates a severe capital lockup problem. If a trader locks $100,000 in an election arbitrage in July for an event resolving in November, that capital is inaccessible for four months. The opportunity cost of this lockup must be measured against the risk-free rate. If short-term Treasury bills yield 5 percent annualized, locking capital for a third of the year requires a minimum return of 1.66 percent just to break even with passive fixed-income investments. ## What does a reconstructed arbitrage P&L look like? To illustrate the mechanics, we can reconstruct a hypothetical trade based on general pricing anomalies. Assume a trader spots a divergence in a hypothetical "Candidate A to win the Election" market. On Kalshi, Candidate A YES is trading at $0.45. On Polymarket, Candidate A NO is trading at $0.50. Buying one YES contract and one NO contract guarantees a payout of $1.00 at resolution, regardless of who wins the election. Here is the step-by-step breakdown for a hypothetical 10,000-contract position: | Metric | Kalshi (Leg 1) | Polymarket (Leg 2) | Total / Net | | :--- | :--- | :--- | :--- | | Position | 10,000 YES | 10,000 NO | Hedged | | Execution Price | $0.45 | $0.50 | $0.95 combined | | Gross Cost | $4,500.00 | $5,000.00 | $9,500.00 | | Taker Fees (Unrounded) | $173.25 | $100.00 | $273.25 | | Total Capital Deployed | $4,673.25 | $5,100.00 | $9,773.25 | | Guaranteed Payout | $10,000.00 | $0.00 (or vice versa) | $10,000.00 | | Net Profit | N/A | N/A | $226.75 | | Return on Capital | N/A | N/A | 2.3201% | *(Note: This case uses a hypothetical CURRENT-fee scenario, not a documented 2024 trade, as 2026 fees cannot be applied as historical 2024 fees. Kalshi unrounded fee for 10,000 contracts at $0.45 is `10000 * 0.07 * 0.45 * 0.55 = 173.25`, not $174 from multiplying rounded 100-lot fees. Polymarket fee for 10,000 contracts at $0.50 with a 0.04 politics rate is `10000 * 0.04 * 0.50 * 0.50 = 100.00` USDC.)* In this hypothetical scenario, the gross spread of 5 cents ($500 on 10,000 contracts) is immediately reduced by $273.25 in unrounded fees. The net profit before rounding of $226.75 represents a 2.3201% return on the $9,773.25 deployed. If settlement takes longer, the annualized equivalent falls. For illustration, the simple annualized equivalent of this modeled return is about 28.23% over 30 days or 4.70% over 180 days, using a 365-day year. Neither figure is a promised repeatable yield; compare the actual holding period, current alternatives, and venue risks. This is not financial advice, but a mathematical reality of market neutral strategies. ## Frequently asked questions ### What happens if a prediction market platform halts trading? A halt or outage can prevent execution, exits, or withdrawals. Settlement timing and recovery depend on the venue, contract rules, and nature of the failure; a quoted payout is not a guarantee that funds remain accessible or will be recovered on schedule. ### Can I use margin to trade prediction market arbitrage? Regulated prediction markets in the United States generally do not offer margin trading for retail participants. All positions must be fully collateralized upfront, which limits the return on equity compared to traditional leveraged financial arbitrage. ### Do maker rebates offset taker fees in cross-venue trades? Yes, if you provide liquidity by resting limit orders rather than crossing the spread, you can avoid taker fees and potentially earn rebates. However, resting orders expose you to execution risk, as the price gap may close before both legs of your arbitrage are filled. Atomic cross-venue execution is not possible. ## Sources * [SSRN: Price Discovery and Trading in Modern Prediction Markets](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5331995) * [Arxiv: The Anatomy of a Blockchain Prediction Market: Polymarket in the 2024 U.S. Presidential Election](https://arxiv.org/html/2603.03136v2) * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) --- ### Market Making vs Arbitrage in Prediction Markets Canonical: https://predictionmarketarbitrage.com/blog/market-making-vs-arbitrage-prediction-markets Published: 2026-10-25 Updated: 2026-11-16 Author: PMA Research Summary: Compare market making and arbitrage in prediction markets. Learn how to capture spreads, earn liquidity rebates, and calculate hybrid strategy profitability. Prediction markets offer two distinct paths for quantitative traders: capturing the spread as a liquidity provider or hunting cross-venue discrepancies as an arbitrageur. While taking risk-free arbitrage looks appealing on paper, the reality of taker fees often destroys the edge. Market making can capture spreads and may qualify for rebates, but it also carries inventory, adverse-selection, and execution risks; it is not inherently profitable or superior to arbitrage. ## Key takeaways * Pure cross-venue taker arbitrage can become unprofitable when fees on both legs exceed the executable spread. Fill-or-Kill (FOK) orders control fills on one venue but never provide atomic cross-venue execution. * Market makers profit by capturing the bid-ask spread and earning venue-specific liquidity incentives rather than taking directional bets. Note that a limit order can be a taker if it is marketable; only resting, non-marketable orders provide liquidity. * Polymarket rewards makers with daily rebates funded by taker fees, with category-specific rebate-pool shares rather than a fixed percentage of trade value for filled resting orders in certain categories. * A hybrid strategy resting a maker order on one venue and hedging as a taker on another can yield positive expected value if execution risks are managed. ## How does market making work in prediction markets? Market making involves simultaneously posting a bid to buy and an offer to sell on the same contract. The gap between these two prices is the spread. When a trader buys from your offer and another sells to your bid, you capture this spread as profit. You provide immediate liquidity to the market, and the spread compensates you for the inventory risk you assume. Inventory risk occurs when the market moves in one direction, leaving you holding a large position of YES or NO contracts. If news breaks, informed traders will cross the spread and fill your resting quotes before you can cancel them. This is known as adverse selection. To survive, market makers must constantly adjust their quotes based on order book dynamics and external probability signals. ## What are the Polymarket maker rebates and liquidity rewards? Polymarket charges taker fees on most categories, while makers pay no fee and can earn rebates. The venue operates a Maker Rebates Program that pays daily USDC rewards to liquidity providers whose resting orders are filled. According to the official documentation, rebates are performance-based and distributed using a fee-curve weighted formula. For each filled maker order, the fee equivalent is calculated based on the number of shares, the fee rate, and the price using the formula `fee = C * feeRate * p * (1 - p)`. Makers earn a percentage of this fee equivalent, which varies by category. For example, Crypto markets offer a 20% rebate, Sports markets offer a 15% rebate, while Finance, Politics, and Economics offer 25% [Polymarket Documentation: Maker Rebates Program](https://docs.polymarket.com/market-makers/maker-rebates). To qualify, your orders must add liquidity to the book and be taken by another trader. Rebates are paid daily directly to your wallet, requiring a minimum accrued balance of $1 USDC [Polymarket Help Center: Maker Rebates Program](https://help.polymarket.com/en/articles/13364471-maker-rebates-program). Markets with a 0% taker fee, such as Geopolitics, do not generate maker rebates. It is important to note that Polymarket fees are rounded to five decimal places in USDC, not cents. ## How do Kalshi fees impact market makers? Kalshi utilizes a different fee structure. The general taker fee formula is `round up(M * 0.07 * C * P * (1 - P))`, where the default taker multiplier (M) is 1. For makers, the formula is `round up(M * 0.0175 * C * P * (1 - P))`. While the default maker multiplier is 0 (meaning no maker fees), there are many series exceptions where maker fees apply [Kalshi Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf). Traders must explicitly acknowledge that current precision must be read from live fee parameters. You should not assume a per-contract 1.75-cent hard cap after rounding or across multipliers. For instance, the general fee for 100 contracts at $0.51 is an unrounded $1.7493, though tables may display about $1.75. There is no universal blanket maker-free claim on Kalshi. ## Why does pure cross-venue arbitrage fail after fees? Arbitrageurs look for situations where the implied probability of an event differs across venues. If a YES contract pays $1.00 at resolution, buying YES at 40 cents on Polymarket and NO at 55 cents on Kalshi seems to lock in a 5-cent gross profit. However, executing this trade requires crossing the spread as a taker on both venues. When you subtract the taker fees for both legs, the net profit can become negative when fees exceed the available margin. Furthermore, venues can resolve the "same" event differently based on their specific rulebooks, introducing settlement mismatch risk. Polymarket uses the UMA optimistic oracle for prediction markets and Chainlink TWAP for up-down markets. UMA challenges take 2 hours, and disputes can take days. Cross-venue opposite labels alone do not guarantee a payout. ## Can a hybrid maker-taker strategy generate consistent profit? Since pure taker arbitrage is structurally unprofitable, quantitative traders often deploy a hybrid approach. This involves resting a maker order on one venue to capture the spread and earn rebates, then immediately hedging as a taker on another venue when the maker order fills. Consider a strictly hypothetical worked example for a political market (Polymarket fee rate 0.04, rebate 25%; Kalshi taker M=1): | Venue | Action | Price | Fee | Rebate | Net Cash Flow | | :--- | :--- | :--- | :--- | :--- | :--- | | Polymarket | Maker Sell YES | $0.520 | $0.00000 | $0 assumed | +$0.52000 | | Kalshi | Taker Buy YES | $0.500 | -$0.01750 | $0.00000 | -$0.51750 | | **Total** | **Hypothetical matched exposure** | **N/A** | **N/A** | **N/A** | **+$0.00250 before Kalshi rounding** | *Note: All fills, timestamps, and prices in this table are hypothetical and do not represent historical reconstruction or guaranteed profitability.* In this scenario, you rest an offer to sell YES at 52 cents on Polymarket. When a taker fills your order, you pay zero fees and earn a fraction of a cent in maker rebates. Your system immediately sends a taker buy order to Kalshi at 50 cents. You pay the Kalshi taker fee. Your net position is flat, and your net profit is half a cent per contract. This strategy relies heavily on execution speed. If the Kalshi price moves before your hedge executes, you are left with unhedged directional exposure. ## API Data Structures for Market Makers When building automated systems, understanding the API data structures is critical. For example, Kalshi's public REST market data requires no authentication, while WebSocket connections do. In the REST API, the `Orderbook_fp.yes_dollars` and `no_dollars` arrays contain dollar prices and fixed-point sizes, sorted ascending with the best bid last. Crucially, only bids are returned; to find the YES ask, you must calculate `1 - best NO bid` [Kalshi API Documentation](https://docs.kalshi.com/getting_started/quick_start_market_data). Missing asks are unavailable, never zero. ## Frequently asked questions ### Do I need special software to be a market maker? Yes. Manual market making is highly difficult due to the speed required to update quotes and manage inventory. You will need a custom trading system connected to the venue APIs. ### How are Polymarket maker rebates paid? Rebates are calculated daily and paid directly to your wallet in USDC. You must accrue a minimum of $1 USDC in rebates to trigger a payout for that day. ### Are all deposits on PredictIt free? No. While PredictIt removed its 5% withdrawal fee on September 23, 2026, the 10% trading profit fee remains. ACH deposits are free, but subsequent credit card deposits incur a 3% fee, though there is a one-time next-card waiver up to $1000. Additionally, the current investment cap is $3500 per contract (updated July 15, 2025, from the historical $850 limit) [PredictIt Product Update](https://www.predictit.org/stories/detail/102/Product-Update--Deposits-and-Withdrawals-are-Free). ## Sources * [Polymarket Documentation: Maker Rebates Program](https://docs.polymarket.com/market-makers/maker-rebates) * [Polymarket Help Center: Maker Rebates Program](https://help.polymarket.com/en/articles/13364471-maker-rebates-program) * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) * [Kalshi API Documentation](https://docs.kalshi.com/getting_started/quick_start_market_data) * [PredictIt Product Update](https://www.predictit.org/stories/detail/102/Product-Update--Deposits-and-Withdrawals-are-Free) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) --- ### Build a Python Cross-Venue Price Monitor for Polymarket and Kalshi Canonical: https://predictionmarketarbitrage.com/blog/python-cross-venue-price-monitor Published: 2026-10-24 Updated: 2026-11-16 Author: PMA Research Summary: Learn how to build a read-only Python price monitor for Polymarket and Kalshi. Normalize prices, map markets, and calculate fee-adjusted spreads. A read-only price monitor is a useful first project for studying cross-venue prediction markets. It can compare actual purchase prices without connecting a wallet or placing orders. It cannot prove that two contracts settle identically or that quotes will still exist when an order arrives. This guide builds an educational snapshot checker with Python's standard library and makes those limitations explicit. ## Key takeaways * Use executable asks for the outcomes you would actually buy, not last-trade prices or midpoints. * Kalshi's order-book response supplies bids. A best NO bid implies a YES ask of one dollar minus that bid. * Polymarket outcome identifiers are decimal integers, not Ethereum wallet addresses. Current discovery documentation distinguishes CTF token identifiers from Protocol V2 position identifiers. * Compare quantities and include both legs' purchase costs and fees. A positive model result is not an executable or guaranteed profit. ## What should the mapping contain? Start by manually reviewing one event on each venue. Record the exact Kalshi ticker and the Polymarket trading identifier for the opposing outcome. A mapping table should contain the Kalshi ticker, Polymarket NO asset identifier, source agency, deadline, and review date. Fill those fields with identifiers from the real markets you have checked; do not copy an invented ticker or a wallet address from a tutorial. Market titles are insufficient. Compare the outcome definition, observation period, time zone, official source, cancellation rules, and unusual-resolution provisions. If one venue's YES and the other's NO can both lose, the pair is not a simple matched hedge. A human-reviewed mapping still needs maintenance when rules change. ## How do the public data interfaces differ? Kalshi documents a public market-data API under its v2 Trade API. Its order-book response uses an orderbook_fp object with dollar-denominated price strings, including no_dollars. Because a NO bid at $0.60 corresponds to an implied YES offer at $0.40, the highest positive-size NO bid determines the best implied YES ask. An empty opposite book means no observable ask, not a zero-cost contract. Polymarket separates market discovery from order-book reads. Gamma discovery data identifies the outcomes; the trading API returns the book for the selected asset identifier. Current documentation distinguishes clobTokenIds for CTF markets and positionIds for Protocol V2 markets. Follow the market's version and outcome mapping rather than assuming the first identifier is always YES. The documented order-book interface accepts the selected trading identifier as token_id. Current Polymarket documentation describes bids in ascending order and asks in descending order. The example does not depend on that ordering: it explicitly selects the minimum positive-size ask. It uses Decimal because converting decimal money strings to binary floating point can introduce avoidable rounding artifacts. ## What does a minimal snapshot checker look like? This example uses public reads only. Run it locally, provide a real reviewed Kalshi ticker and the opposing Polymarket NO identifier, and choose a quantity no larger than the displayed best levels. Fee inputs are assumptions to verify against the actual market, not values discovered automatically by this script. ~~~python import json import urllib.request from urllib.parse import quote, urlencode from decimal import Decimal, ROUND_HALF_UP D = Decimal def get_json(url): req = urllib.request.Request(url, headers={"User-Agent": "read-only-research/1.0"}) with urllib.request.urlopen(req, timeout=10) as response: return json.load(response) ticker = input("Reviewed Kalshi ticker: ").strip() asset_id = input("Reviewed Polymarket NO asset identifier: ").strip() if not ticker or not asset_id.isdecimal(): raise ValueError("Supply a ticker and a decimal outcome identifier.") kalshi_url = "https://external-api.kalshi.com/trade-api/v2/markets/" kalshi = get_json(kalshi_url + quote(ticker, safe="") + "/orderbook") poly_url = "https://clob.polymarket.com/book?" + urlencode({"token_id": asset_id}) poly = get_json(poly_url) no_bids = [(D(x[0]), D(x[1])) for x in kalshi["orderbook_fp"]["no_dollars"] if D(x[1]) > 0] asks = [(D(x["price"]), D(x["size"])) for x in poly["asks"] if D(x["size"]) > 0] if not no_bids or not asks: raise ValueError("A required book side has no available liquidity.") best_no = max(price for price, size in no_bids) k_price = D("1") - best_no p_price = min(price for price, size in asks) k_size = sum((size for price, size in no_bids if price == best_no), D("0")) p_size = sum((size for price, size in asks if price == p_price), D("0")) quantity = D(input("Whole-contract quantity: ")) if quantity <= 0 or quantity != quantity.to_integral_value(): raise ValueError("Use a positive whole-contract quantity in this example.") if quantity > min(k_size, p_size): raise ValueError("Quantity exceeds a displayed best-price level.") k_multiplier = D(input("Verified Kalshi taker multiplier: ")) p_rate = D(input("Verified Polymarket fee coefficient: ")) if k_multiplier < 0 or p_rate < 0: raise ValueError("Fee inputs must be nonnegative.") k_raw = k_multiplier * D("0.07") * quantity * k_price * (D("1") - k_price) p_raw = quantity * p_rate * p_price * (D("1") - p_price) p_fee = p_raw.quantize(D("0.00001"), rounding=ROUND_HALF_UP) purchases = quantity * (k_price + p_price) modeled_cost = purchases + k_raw + p_fee print("Kalshi YES ask:", k_price, "Polymarket NO ask:", p_price) print("Modeled cost before Kalshi rounding:", modeled_cost) print("Matched payout minus modeled cost:", quantity - modeled_cost) ~~~ The script intentionally stops on missing or unexpected data instead of substituting a zero price. HTTP failures, unavailable identifiers, and schema changes need investigation. Never reinterpret an error response as an opportunity. No credentials or signing keys are required for these documented public reads. ## How should the fee output be interpreted? Kalshi's general taker formula is M × 0.07 × quantity × price × (1 − price), with the applicable multiplier and rounding rules. The checker reports the raw formula amount, not an exact invoice. Use the current market-specific schedule and actual fill records to implement production fee precision. Polymarket's documented taker fee is quantity × feeRate × price × (1 − price), rounded to five decimal places in USDC. Category coefficients differ, and the selected market's actual parameters control. Maker treatment does not apply merely because an order is a limit order: an immediately marketable limit order can take liquidity. For a hypothetical 100-pair purchase at Kalshi $0.40 and Polymarket $0.55, purchases total $95. At a Kalshi multiplier of 1 and a hypothetical Polymarket coefficient of 0.04, fees are $1.68 raw and $0.99000 respectively. The modeled total is $97.67 and the matched $100 payout leaves $2.33 before Kalshi rounding and other costs. These inputs are illustrative, not observed quotes. ## Why is this not a production arbitrage bot? The HTTP requests happen sequentially, so the snapshots are not simultaneous. Displayed liquidity can cancel, prices can move, and the script does not walk deeper levels or send orders. It does not automatically verify contract equivalence, detect every stale snapshot, discover fees, or account for funding and withdrawal costs. Its output is a candidate for investigation. For larger quantities, compute depth-weighted costs by consuming available levels and applying fees to the relevant fills. For continuous monitoring, consult current rate limits, use bounded requests and backoff after throttling, and consider documented streaming feeds. No arbitrary polling frequency is universally safe. A Fill-or-Kill instruction on one venue cannot make two separate venues execute atomically. ## Frequently asked questions ### Do I need private keys for this monitor? No. Keep it read-only. Review our [fake bot checklist](/fake-bot-checklist) before running unfamiliar software or installing dependencies. ### Is the displayed result guaranteed profit? No. It omits Kalshi venue rounding and other frictions, and requires complementary settlement. Use our [fee calculator](/fee-calculator) to explore sensitivity, not to replace venue terms. ### Can blockchain logs reconstruct the historical order book? Executed on-chain trades can validate some fills. They do not reveal every resting off-chain order or cancellation. Historical depth requires an appropriate archive. ## Sources * [Kalshi: Quick Start, Market Data](https://docs.kalshi.com/getting_started/quick_start_market_data) * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) * [Polymarket: Prices and Order Books](https://docs.polymarket.com/market-data/prices-order-books) * [Polymarket: Discover Markets](https://docs.polymarket.com/market-data/discover-markets) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) --- ### Polymarket vs Kalshi: Fees, Resolution Rules and Liquidity Canonical: https://predictionmarketarbitrage.com/blog/polymarket-vs-kalshi-fees-rules-liquidity Published: 2026-10-23 Updated: 2026-11-16 Author: PMA Research Summary: A deep comparison of Polymarket and Kalshi for arbitrageurs, covering regulatory status, fee models, UMA oracle mechanics, and liquidity. Arbitrage across prediction markets looks like free money until a resolution mismatch or hidden fee wipes out your margin. For quantitative traders bridging Polymarket and Kalshi, success requires mastering the structural differences in regulatory status, oracle mechanics, and settlement rails. ## Key takeaways * Kalshi operates as a fully CFTC-regulated Designated Contract Market, while Polymarket splits into an international crypto platform and a separate US entity. * Both platforms utilize a parabolic taker fee model peaking at 50 cents, but Polymarket's rates vary by category and are calculated to five decimal places in USDC. * Kalshi resolves markets centrally via its rulebook and markets team, whereas Polymarket international relies on the UMA optimistic oracle for prediction markets and Chainlink TWAP for up-down markets. * Arbitrageurs must account for deposit rail friction, comparing Kalshi's fiat USD wires against Polymarket's Polygon-based USDC infrastructure. ## How do Kalshi and Polymarket differ in regulatory status? Regulatory architecture dictates capital efficiency and market access. Kalshi operates as a Designated Contract Market regulated by the Commodity Futures Trading Commission. This status allows it to offer event contracts directly to US residents, settling entirely in US dollars. The platform requires standard identity verification and operates within the traditional financial system's compliance frameworks. Note that CFTC regulation is not FDIC insurance or a guarantee of risk-free capital. Polymarket presents a bifurcated structure. The primary international platform operates offshore and strictly geoblocks US residents. It functions entirely onchain, utilizing smart contracts on the Polygon network. To serve the US market, a separate entity known as Polymarket US operates as its own CFTC-designated contract market. Arbitrageurs must maintain distinct legal and technical infrastructure to interact with these separate liquidity pools. You cannot simply route orders between them using the same API credentials or capital base. ## What are the fee models for Kalshi and Polymarket? Fees dictate the viability of any statistical arbitrage strategy. Both venues employ a parabolic fee structure that extracts the highest toll when uncertainty is greatest at the 50-cent mark. Any fee assumptions must match the specific venue and market. The standard Kalshi taker fee formula is `round up(M * 0.07 * C * P * (1 - P))`, where M is the multiplier (defaulting to 1 for takers), C is the number of contracts, and P is the price. This maxes out at about 1.75 cents per contract at 50 cents, though current precision must be read from live fee parameters as rounding rules apply to the centicent level. Polymarket charges taker fees based on the formula `fee = C * feeRate * p * (1 - p)`, rounded to five decimal places in USDC. According to the official [Polymarket: Trading Fees](https://docs.polymarket.com/trading/fees) documentation, makers are never charged fees, and the taker fee rate varies by category: 0.07 for Crypto, 0.05 for Sports/Economics/Culture/Weather/Other, 0.04 for Finance/Politics/Mentions/Tech, and 0 for Geopolitics. ### Worked numeric example Consider a hypothetical arbitrage trade where you buy 10,000 YES contracts at 45 cents on Kalshi and 10,000 NO contracts at 50 cents on Polymarket (Politics category). A YES contract typically pays $1.00 at resolution, though rare Unknown/50-50 outcomes can redeem at $0.50. * **Kalshi Taker Fee (Raw before rounding):** `1 * 0.07 * 10,000 * 0.45 * (1 - 0.45)` = $173.25. * **Polymarket Taker Fee (Politics):** `10,000 * 0.04 * 0.50 * (1 - 0.50)` = $100.00000 USDC. Net profit before venue rounding is $10,000 - $9,500 - $173.25 - $100 = $226.75. The required capital is $9,773.25, yielding a return of 2.3201%. If you provide liquidity as a maker on Polymarket, your fee is zero, and you may even earn a rebate, though rebate eligibility is not guaranteed revenue. | Platform | Category | Taker Coefficient | Maker Fee | Maker Rebate | | :--- | :--- | :--- | :--- | :--- | | Kalshi | Standard | 0.07 | M * 0.0175 | None | | Polymarket | Crypto | 0.07 | 0 | 20% | | Polymarket | Sports | 0.05 | 0 | 15% | | Polymarket | Finance | 0.04 | 0 | 25% | ## UMA optimistic oracle vs Kalshi rulebook: Who decides the outcome? Arbitrageurs face severe basis risk because venues can resolve the "same" event differently. Cross-venue opposite labels alone do not guarantee a payout. Kalshi relies on a centralized resolution process. The exchange decides results using the specific source agencies named in each contract's rulebook. Polymarket international utilizes a decentralized approach via the UMA optimistic oracle for standard prediction markets. As detailed in the [Polymarket: Resolution](https://docs.polymarket.com/concepts/resolution) documentation, anyone can propose a resolution, initiating a 2-hour challenge period. If disputed, the resolution escalates to the Data Verification Mechanism, where UMA token holders vote. This process can delay settlement by days. ## How do deposit rails and API access compare? Kalshi operates on traditional fiat rails. Deposits and withdrawals occur via ACH or wire transfers. Kalshi's REST API provides market data, but responses are cached and can lag seconds. To calculate the YES ask, you must use `1 - best NO bid` from the `orderbook_fp.no_dollars` array, as only bids are returned. ```python import urllib.request import json # Educational example: Fetching Kalshi REST orderbook (asynchronous/stale) def get_kalshi_yes_ask(ticker): url = f"https://external-api.kalshi.com/trade-api/v2/markets/{ticker}/orderbook" req = urllib.request.Request(url) try: with urllib.request.urlopen(req) as response: data = json.loads(response.read()) no_bids = data.get("orderbook_fp", {}).get("no_dollars", []) if not no_bids: return None best_no_bid = no_bids[-1][0] # Sorted ascending, best bid last return 1.0 - best_no_bid except Exception as e: return None ``` Polymarket international operates entirely on crypto rails using USDC on the Polygon network. API access requires wallet signatures. ## How does typical liquidity impact arbitrage execution? Liquidity dictates slippage. Remember that a limit order can be a taker if it is marketable; only resting, non-marketable orders provide liquidity. Furthermore, Fill-or-Kill (FOK) controls fills on one venue only; there is no atomic cross-venue FOK execution. Polymarket incentivizes liquidity through its maker rebate program. By redistributing a portion of taker fees to makers, Polymarket encourages tighter spreads. However, this liquidity requires specialized onchain execution logic. ## Frequently asked questions ### Can US residents trade on Polymarket? US residents are strictly prohibited from trading on the international Polymarket platform. They must use the separate, CFTC-regulated Polymarket US entity. ### How long does a disputed Polymarket resolution take? If a proposed outcome is challenged during the initial 2-hour window, the dispute goes to a vote by UMA token holders, which can take days to finalize. ### Do maker orders pay fees on Kalshi? There is no universal blanket maker-free claim on Kalshi. The default maker multiplier is M=0, but many series exceptions exist where M=1, resulting in a maker fee formula of `M * 0.0175 * C * P * (1 - P)`. ## Sources * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) * [Polymarket: Resolution](https://docs.polymarket.com/concepts/resolution) * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) --- ### How Prediction Market Arbitrage Scanners Work (and Where They Mislead) Canonical: https://predictionmarketarbitrage.com/blog/how-arbitrage-scanners-work Published: 2026-10-21 Updated: 2026-11-16 Author: PMA Research Summary: Inside prediction market arbitrage scanners: how market matching, depth-weighted pricing, and fee-adjusted spreads separate real edge from false positives. A quoted gap is only the beginning of the check. To find real edge, a prediction market arbitrage scanner must verify matching outcomes, executable depth, and fees across venues. Here is how the tools actually work, and why the numbers they report are so often wrong. ## Key takeaways * Some scanners compute gaps using midpoints, but midpoints are not executable asks; you must buy at the actual ask and sell at the bid. Both directions of an arbitrage trade use actual BUY asks on opposite sides. * Cross-venue matching is entity resolution on prose, not a simple ticker lookup, making it an important source of false positives. * A gap that clears at 10 contracts often does not clear at 500, which is why depth-weighted pricing is required to calculate real edge. * Two contracts can describe the same event but settle differently on an edge case, leaving you with two losing positions instead of a hedge. Cross-venue opposite labels alone do not guarantee payout. ## How do scanners match markets across venues? Equities have a symbol. Bonds have an ISIN. Prediction markets have whatever the venue called the event that morning. Because there is no shared identifier for a real-world event, establishing that two markets describe the same thing is a matching problem with real error rates. The same rate decision might appear on Polymarket with a conversational title, and on Kalshi with a title written like a contract clause. A naive scanner relies on fuzzy title matching. A useful scanner builds an explicit contract identity record, mapping the event definition, threshold, deadline, time zone, and resolution authority. If a rule cannot be checked, the scanner should mark the pair for human review rather than silently treating it as a match. A false match produces a scanner that confidently reports arbitrage between two contracts that will settle differently. For example, Polymarket uses the UMA optimistic oracle for prediction markets (where disputes can take days) and Chainlink TWAP for up-down markets. Rules define resolution, not the title. ## Why do scanners report gaps that do not exist? The most common reason scanners report phantom edge is that they compute on midpoints. A midpoint is the average of the best bid and the best ask. It is a convenient summary, but it is not a price anyone will trade with you. You buy at the ask and sell at the bid, so a strategy priced on midpoints has already given away half the spread on each leg before it starts. Missing asks are unavailable, never zero. There is a second illusion stacked on the first: top-of-book pricing. Even the best ask is only the top of the book. If there are forty contracts available at that price and you want five hundred, you will walk up the ladder and pay worse on the rest. A scanner that reports one number without a size attached has not answered the question you asked. This is why size has to be an input rather than an afterthought, and why depth-aware pricing is mandatory. ## How do fees impact the spread? A price difference becomes actionable only after it survives fee deductions on both legs. Both major venues charge taker fees that scale with the probability of the contract. A limit order can be a taker if it is marketable; only resting/non-marketable orders provide liquidity and act as makers. For example, the [Kalshi taker fee](https://kalshi.com/docs/kalshi-fee-schedule.pdf) is calculated using a general taker formula: `round up(M * 0.07 * C * P * (1 - P))`, where the default taker M=1. The July 7, 2026 effective schedule states that the fee plus position cost is rounded up to a centicent, though printed fee tables often round to cents. Explicitly, current precision must be read from live fee parameters. [Polymarket expanded taker fees](https://docs.polymarket.com/trading/fees) to nearly every category on 30 March 2026. The fee is `C * feeRate * p * (1 - p)`. Rates vary: Crypto is 0.07, Sports/Economics/Culture/Weather/Other is 0.05, Finance/Politics/Mentions/Tech is 0.04, and Geopolitics is 0. Polymarket fees are rounded to five decimal places in USDC, not cents. Makers are never charged fees on Polymarket, though rebate eligibility is not guaranteed revenue. ### Worked Example: Fee-Adjusted Spread Assume a standard winning token pays $1.00 and a losing token pays $0.00 at resolution. You want to buy 100 YES on Kalshi and 100 NO on Polymarket (Politics category). | Metric | Kalshi (YES) | Polymarket (NO) | |---|---|---| | Depth-weighted Ask (100 qty) | $0.51 | $0.46 | | Taker Fee Rate | 0.07 | 0.05 | | Raw Fee (100 qty) | $1.7493 | $1.24200 | *Hypothetical scenario for illustration only. Return calculations must include both legs' capital and costs.* Without fees, the spread looks like a $0.03 profit per contract ($1.00 - $0.51 - $0.46). With depth-weighted pricing and fees, the raw fees before venue rounding are $1.7493 on Kalshi and $1.24200 on Polymarket. The total capital required is $99.9913, leaving a net profit of $0.0087 before venue rounding. ## What causes false positives in live feeds? Beyond midpoints and top-of-book illusions, scanners generate false positives through stale quotes and reconstructed depth. If one venue's price is thirty seconds old and the other is live, the gap you are seeing may be a timing artifact. A one-minute polling interval can miss quote changes between reads; snapshot age and endpoint behavior must be measured rather than assumed. Serious scanners consume WebSocket feeds (which require authentication on platforms like Kalshi) to maintain a live local order book. Cheap contracts also produce systematic false positives. A one-cent spread on a three-cent contract is enormous in percentage terms and almost always unfillable. Longshots are where naive scanners find their best-looking opportunities and where those opportunities are least real. Finally, beware of reconstructed depth. On some venues, the book is reconstructed from a pricing curve. If your scanner treats reconstructed depth as real resting orders, it will report size that nobody will fill. ## How should you evaluate a scanner's claims? When evaluating a scanner, look for what it does when it is unsure. A scanner should fail closed. When a gate cannot be evaluated—such as resolution equivalence—it should say so rather than upgrading the row anyway. "We could not verify this" and "we verified there is no edge" are completely different statements. Data freshness, depth coverage, fee handling, and API availability vary by provider and plan. Verify each capability directly rather than inferring it from whether a service is paid. However, no scanner solves fill certainty or capital sitting on the wrong venue. Furthermore, be extremely cautious of open-source bots; fake arbitrage bot repos have stolen private keys, and security VM/burner mitigations are not safety guarantees. Always use a fee calculator to verify the math yourself. *(Note: This is not financial advice. Arbitrage carries execution and smart contract risks. CFTC regulation is not FDIC insurance or risk-free capital.)* ## Frequently asked questions ### Do scanners execute the trades automatically? Most commercial scanners are read-only tools that surface opportunities. Execution requires a separate bot or manual trading. FOK (Fill-or-Kill) controls fills on ONE venue, never atomic cross-venue. Two fill-or-kill orders do not make a two-leg trade atomic. ### Why do scanners show arbitrage on markets that are already closed? This is a late-resolution gap. One venue may have settled the market while the other still allows trading. The scanner sees a price discrepancy, but you cannot actually buy the settled side. ### Can two markets have the exact same title but settle differently? Yes. Venues can resolve the "same" event differently based on their specific rulebooks, such as using different official sources or cutoff times. This is why resolution equivalence is a mandatory check. ## Sources * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) * [Kalshi Docs: Quick Start Market Data](https://docs.kalshi.com/getting_started/quick_start_market_data) * [Polymarket Docs: Resolution](https://docs.polymarket.com/concepts/resolution) * [CFTC: Prediction Markets](https://www.cftc.gov/LearnandProtect/PredictionMarkets) --- ### Cross-Venue Arbitrage Between Polymarket and Kalshi Playbook Canonical: https://predictionmarketarbitrage.com/blog/cross-venue-arbitrage-playbook Published: 2026-10-21 Updated: 2026-11-16 Author: PMA Research Summary: Master cross-venue arbitrage between Polymarket and Kalshi with this practical playbook covering fee math, order types, and resolution risks. Cross-venue arbitrage between Polymarket and Kalshi looks like guaranteed profit on a screen, but the reality involves navigating complex fee structures, execution mechanics, and settlement risks. Executing a profitable spread requires strict discipline in matching resolution criteria, sizing both legs correctly, and avoiding execution slippage. This playbook breaks down the exact mechanics of locking in a return when prices diverge across these two major prediction markets. It is important to remember that while these markets are regulated or operate under specific frameworks, CFTC regulation is not FDIC insurance or risk-free capital. ## Key takeaways * A standard winning event contract usually pays $1.00 at resolution, but rare Unknown or 50-50 outcomes can redeem at $0.50. Cross-venue opposite labels alone do not guarantee a full payout. * Kalshi taker fees use a formula that scales with a multiplier, meaning there is no universal 1.75-cent hard cap per contract across all markets. * Polymarket taker fees vary by category (e.g., 0.04 for politics, 0.07 for crypto) and are rounded to exactly five decimal places in USDC, not cents. * Venues frequently resolve seemingly identical events differently, making rigorous rulebook comparison mandatory. * A limit order can be a taker if it is marketable; only resting, non-marketable orders provide liquidity and qualify for maker fee treatment. ## How do you match equivalent markets across venues? The foundation of prediction market arbitrage is ensuring that the two contracts you are trading actually represent the exact same outcome. A standard YES contract typically pays $1.00 at resolution, but only if the underlying event triggers the specific settlement criteria defined by the exchange. Venues can resolve the same event differently based on their individual rulebooks. Before committing capital, you must compare the official resolution sources. If Kalshi uses a specific government press release to determine an economic data point, Polymarket must use that exact same press release. If one uses a consensus of news networks for an election call and the other uses official state certification, you do not have an arbitrage opportunity. You have a directional bet on the timing and agreement of those two sources. Always read the fine print. Furthermore, the oracle mechanisms differ. Polymarket relies on the UMA optimistic oracle for prediction markets and Chainlink TWAP for up-down markets. The UMA oracle involves a 2-hour challenge period, and disputes can take days to resolve. Rules define the resolution, not the market title. In rare cases, markets may resolve as Unknown or 50-50, redeeming each share at $0.50. Therefore, holding opposite labels across venues does not universally guarantee a $1.00 payout. ## What are the exact fee formulas for Kalshi and Polymarket? Gross margins are irrelevant if fees consume your entire edge. Both platforms charge fees based on the size and probability of the trade, and these formulas dictate your actual entry thresholds. Any fee assumptions must match the specific venue and market. Kalshi calculates its trading fees using a specific formula. The general Kalshi taker fee is calculated as `round up(M * 0.07 * C * P * (1-P))`, where M is the multiplier (default taker M=1), C is the number of contracts, and P is the price. The July 7, 2026 effective schedule states that the fee plus position cost is rounded up to a centicent, though printed fee tables often round to cents. You must explicitly read the current precision from live fee parameters. There is no universal blanket maker-free claim on Kalshi; the maker formula is `M * 0.0175 * C * P * (1-P)`, with the default maker M=0, but there are many series exceptions where M=1 (charging a quarter of the taker rate) or M=2 for certain combos [Kalshi: Fee Schedule for July 2026 - 7.7.26 Update](https://kalshi.com/docs/kalshi-fee-schedule.pdf). Polymarket taker fees follow a similar mathematical curve: `fee = C * feeRate * p * (1 - p)`. The fee rate depends on the category: Crypto is 0.07, sports/economics/culture/weather/other are 0.05, finance/politics/mentions/tech are 0.04, geopolitics is 0, and combos are 0.05. Crucially, Polymarket fees are rounded to five decimal places in USDC, not cents. Makers are never charged fees, and while rebates exist, eligibility does not guarantee revenue [Polymarket: Fees](https://docs.polymarket.com/trading/fees). ## How to calculate fee-adjusted entry thresholds? To determine if a spread is actually profitable, you must calculate the net cost of both legs after applying the formulas above. Return calculations must include both legs' capital and costs. Hypothetical tables are not observed trades. Consider a scenario where you want to buy 1,000 contracts of YES on Kalshi and 1,000 contracts of NO on Polymarket. ### Worked numeric example * Kalshi YES price: $0.42 * Polymarket NO price: $0.55 * Gross cost per pair: $0.97 * Gross profit per pair: $0.03 Now, apply the taker fees for both platforms to see if the trade survives execution. * Kalshi raw taker fee calculation before venue rounding (assuming default M=1): `0.07 * 1000 * 0.42 * 0.58 = 17.052`. * Polymarket taker fee calculation (assuming a 0.04 rate for politics): `0.04 * 1000 * 0.55 * 0.45 = 9.90000` USDC. * Total raw fees: $26.952 * Gross profit on 1,000 pairs: $30.00 * Net profit before venue rounding: $3.048 In this scenario, a seemingly massive 3-cent gap yields a net profit of just over three dollars on nearly a thousand dollars of deployed capital. If you suffered even one cent of slippage on either leg, the trade would instantly become a loss. This highlights why taking liquidity on both sides is rarely a viable long-term strategy. ## Which order types minimize execution risk? Because taker fees destroy the majority of cross-venue arbitrage margins, successful traders rely heavily on maker orders. However, it is critical to understand order semantics: a limit order can be a taker if it is marketable. Only resting, non-marketable orders provide liquidity and qualify for maker fee exemptions or rebates. Using maker orders introduces execution risk. If you post a resting limit order on Kalshi and simultaneously take liquidity on Polymarket, your Kalshi order might not fill. This leaves you with a naked, unhedged position. If the market moves against you, the losses on that single leg will wipe out the profits from dozens of successful arbitrage trades. Traders often use specific time-in-force instructions. Fill-or-Kill (FOK) controls fills on ONE venue, ensuring an order is either completely filled immediately or canceled. However, FOK is never atomic cross-venue. FOK and Immediate-or-Cancel (IOC) semantics differ, and you must understand how each exchange handles them. When querying order books via API, remember that missing asks are unavailable, never zero. Both directions use actual BUY asks. For example, on the Kalshi REST API, only bids are returned; the YES ask is calculated as 1 minus the best NO bid. ## How should you track post-trade settlement? Arbitrage is not complete until the funds are back in your account. Once both legs of the trade are filled, you must monitor the settlement process on both platforms. Keep a detailed ledger recording the exact rulebook text for both contracts at the time of execution, the fill prices, the fees paid, and the expected payout dates. When the event occurs, verify that both platforms resolve the market according to their stated rules. Delays in settlement can tie up your capital, reducing your overall velocity and annualized returns. If a platform disputes a resolution (such as a UMA challenge on Polymarket), you will need your documentation to understand why the payout was delayed. There is no guaranteed profitability, and security mitigations are not absolute safety guarantees. ## Frequently asked questions ### What happens if the two platforms resolve the market differently? If Kalshi and Polymarket resolve the same event with different outcomes, you could lose both legs of your trade. This is why you must verify that the official resolution sources and exact wording match perfectly before executing. Cross-venue opposite labels alone do not guarantee a payout. ### Why do my gross profits disappear after execution? Gross profits do not account for taker fees, which scale with the price and size of your order. Kalshi taker fees use a multiplier formula, and Polymarket charges category-specific rates (e.g., 0.04 or 0.05), which easily consume a 2-cent or 3-cent price gap when taking liquidity on both sides. ### How do I access market data for automation? Public REST market data on Kalshi requires no authentication, while WebSocket connections require auth. REST market responses are cached and can lag by seconds. Always read current endpoint schemas before editing code, as older fields may be deprecated. ## Sources * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) --- ### How Prediction Markets Work: Contracts, Prices and Order Books Canonical: https://predictionmarketarbitrage.com/blog/how-prediction-markets-work-contracts-order-books Published: 2026-10-20 Updated: 2026-11-16 Author: PMA Research Summary: Learn how prediction markets work, from event contracts and order books to price probabilities, resolution rules, and arbitrage fundamentals. Prediction markets strip away the noise of pundits and polls, replacing them with hard financial stakes. By trading event contracts that pay out based on real-world outcomes, participants generate real-time probabilities on everything from elections to economic data. Understanding the mechanics of these exchanges is the first step to finding edge and avoiding costly structural mistakes. ## Key takeaways * Event contracts are financial instruments that typically pay $1.00 if an outcome occurs and zero if it does not, though other structures like ranges exist. * The trading price of a contract acts as the market's implied probability, meaning a 62-cent price reflects a 62 percent chance of occurrence. * Central limit order books match buyers and sellers, with resting orders providing liquidity and marketable orders taking it. * Resolution mechanisms vary drastically by venue, with regulated exchanges relying on internal rulebooks and decentralized platforms using blockchain oracles. ## What are event contracts and how does price equal probability? At their core, prediction markets are exchanges where participants trade claims on future events. Instead of buying shares in a corporation, you buy shares in an outcome. The most common structure is the binary event contract. In a standard binary market, the question is framed as a simple "Yes" or "No" proposition. If the event occurs, the YES contract usually pays exactly $1.00 at resolution. If the event does not occur, the YES contract expires worthless at $0.00. Because the payout is strictly capped at one dollar, the price of the contract naturally behaves like a probability estimate. If a YES contract is trading at $0.62, the price implies 62 percent under the usual simplified interpretation; fees, risk preferences, liquidity, and timing can make price differ from the true event probability. Academic evidence has long supported this mechanism; research on forecasting accuracy demonstrates that market-generated forecasts are typically highly accurate and outperform most moderately sophisticated benchmarks, as detailed in foundational studies on the subject [American Economic Association: Prediction Markets](https://www.aeaweb.org/articles?id=10.1257/0895330041371321). This pricing mechanism is not a poll. It is the result of financial tug-of-war. Traders who believe an outcome is underpriced will buy, pushing the price up. Those who believe it is overpriced will sell, pushing the price down. ## How do order books, bids, and asks function? Prediction markets operate using a central limit order book, identical to traditional equity exchanges. The order book is a live ledger of all resting buy and sell orders at various prices. Participants who post non-marketable orders to the book are "makers" because they provide liquidity. Participants who execute against those resting orders are "takers" because they take liquidity away. The highest price a buyer is willing to pay is the best bid, and the lowest price a seller is willing to accept is the best ask. The gap between them is the spread. When you place a limit order, you specify the exact price you are willing to accept. If it is not immediately marketable, your order will sit on the book until another trader matches it. However, a limit order can also be a taker if it crosses the spread. When you place a market order, you demand immediate execution at the best available price, crossing the spread and paying the taker fee. Consider a worked numeric example for a contract on whether the Federal Reserve will cut interest rates: * You believe the true probability is 65 percent. * The best bid is $0.60 and the best ask is $0.62. * You place a market order to buy 1,000 YES contracts, taking the $0.62 ask. * Your total cost before fees is $620. * If the Fed cuts rates, your 1,000 contracts resolve to $1.00 each, returning $1,000 for a $380 gross profit. Fees heavily impact this math. For instance, the Kalshi general taker fee is calculated as round up(M * 0.07 * C * P * (1 - P)), where M defaults to 1. For 100 contracts at 50 cents, the unrounded fee is $1.75. Polymarket charges taker fees across various categories (e.g., 0.07 for Crypto, 0.05 for Sports/Economics/Culture/Weather/Other, 0.04 for Finance/Politics/Mentions/Tech, and 0 for Geopolitics), calculated as C * feeRate * p * (1 - p) and rounded to five decimal places USDC. Makers on Polymarket are never charged fees and may earn rebates. Always run your expected trades through a fee calculator before executing. ## Glossary of essential prediction market terms To trade effectively, you must speak the language of the exchange. Here is a beginner-friendly breakdown of core concepts. | Term | Definition | Example | | :--- | :--- | :--- | | Event contract | An outcome-linked contract typically paying $1.00 if an outcome occurs and $0.00 otherwise. | "US GDP exceeds 2.5 percent in Q3" | | Implied probability | The contract price read as a percentage. | A $0.75 price implies a 75 percent chance. | | Limit order | An order to buy or sell at a specific price or better. | Bidding $0.40 for 100 YES contracts. | | Market order | An order to execute immediately at the best available current price. | Buying 100 YES contracts at the $0.42 ask. | | Spread | The difference between the highest bid and lowest ask. | Bid $0.40, Ask $0.42 (Spread is $0.02). | ## How do prediction markets resolve and who regulates them? The most critical moment in an event contract's lifecycle is resolution. This is when the market closes and the final outcome is determined. However, the mechanics of resolution depend entirely on the venue. Kalshi operates as a designated contract market regulated by the Commodity Futures Trading Commission (CFTC), having received designation on November 4, 2020. Note that CFTC regulation is not FDIC insurance. When a market is created, Kalshi publishes a strict rulebook defining exactly what constitutes a YES or NO outcome, including the specific data source that will be used as the source of truth. Kalshi acts as the central clearinghouse and resolves the market internally based on those predefined rules. Polymarket operates on the blockchain and relies on a decentralized oracle system known as the UMA protocol for prediction markets, and Chainlink TWAP for up-down markets. To resolve a market via UMA, a participant must propose an outcome and post a bond. If the proposal goes unchallenged during a 2-hour window, the market resolves. If it is disputed, UMA token holders vote to determine the final outcome [Polymarket Help Center: How Are Prediction Markets Resolved?](https://help.polymarket.com/en/articles/13364518-how-are-prediction-markets-resolved). Because resolution rules differ, two venues can resolve the "same" real-world event differently. A trader attempting arbitrage across platforms must read the fine print of both contracts to ensure they are actually trading identical outcomes. ## What are the risks of trading and arbitrage? While prediction markets offer unique opportunities, they carry distinct risks. The most obvious is total loss of capital; if you hold a contract to expiration and the outcome goes against you, your investment goes to zero. This is not financial advice, but position sizing is critical. Arbitrageurs face additional structural risks. Legitimate arbitrage involves buying YES on one exchange and NO on another to lock in a risk-free profit when the combined price is under $1.00. However, execution risk is high. Fill-or-Kill (FOK) orders control fills on one venue but are never atomic cross-venue. If you buy the YES leg but the NO leg moves before your second order fills, you are left with naked directional exposure. Furthermore, the ecosystem is rife with technical hazards. Many newcomers look for automated solutions, but malicious actors prey on this demand. Always verify code and consult a fake bot checklist before deploying third-party software. ## Frequently asked questions ### What happens if a market resolves ambiguously? Exchanges have contingency rules for ambiguous events. Regulated platforms like Kalshi will refer to their specific contract rulebook, which often includes fallback data sources or cancellation clauses. Decentralized platforms may rely on token holder votes to determine if a market should resolve to a specific outcome or be voided. ### Can I sell my contract before the event happens? Yes, provided there is sufficient liquidity in the order book. You can sell your YES or NO contracts to other participants at the current market price at any time before the market officially closes and resolves. ### Why do prices sometimes spike without any news? Prices can move due to liquidity constraints rather than new information. In thinly traded markets, a single large market order can sweep the order book, causing a temporary price spike until market makers step in to provide new liquidity and push the price back to equilibrium. ## Sources * [American Economic Association: Prediction Markets](https://www.aeaweb.org/articles?id=10.1257/0895330041371321) * [Polymarket Help Center: How Are Prediction Markets Resolved?](https://help.polymarket.com/en/articles/13364518-how-are-prediction-markets-resolved) * [Polymarket: Fees](https://docs.polymarket.com/trading/fees) * [Kalshi: Fee Schedule](https://kalshi.com/docs/kalshi-fee-schedule.pdf)