New academic research quantifies the scale of prediction market arbitrage. The finding that matters most: sports markets offer the most opportunities, yet traders are barely touching them.
The Paper
Researchers from IMDEA Networks Institute and the Oxford Internet Institute have published the first large-scale empirical study of arbitrage on Polymarket: "Unravelling the Probabilistic Forest: Arbitrage in Prediction Markets" (Saguillo, Ghafouri, Kiffer & Suarez-Tangil, 2025).
They analysed 86 million bets across thousands of prediction markets between April 2024 and April 2025. The headline finding: an estimated $40 million in realised arbitrage profit was extracted by traders exploiting mispriced bets.
The full paper is available at arxiv.org/abs/2508.03474.
How Prediction Market Arbitrage Works
On Polymarket, every market resolves to either YES or NO. The prices of all possible outcomes should theoretically sum to 100 cents on the dollar, representing a combined 100% probability.
But they frequently don't.
When outcome prices temporarily add up to less than the full dollar, a trader can buy all outcomes and guarantee a profit regardless of what happens. When they add up to more, selling positions locks in risk-free returns. This is textbook arbitrage — the same principle that keeps traditional financial markets efficient.
The researchers identified two distinct types:
Market Rebalancing Arbitrage
This occurs within a single market or condition. Prices drift from their theoretical equilibrium due to uneven buying pressure, and sophisticated traders snap them back. This was the dominant source of profit — buying underpriced NO positions alone accounted for $17.3 million.
Combinatorial Arbitrage
This exploits inconsistencies between logically related markets. If "Team A wins the league" is priced at 60% but "Team A finishes top two" is priced at 55%, there's a logical impossibility being priced into the market. The researchers found 13 exploitable dependent pairs during the U.S. presidential election, with approximately $95,000 extracted.
The Sports Finding
Here's what caught our attention.
Sports markets had the most arbitrage opportunities of any category at accessible budget levels. The paper's analysis at a modest $100 budget threshold showed sports consistently leading in available mispricing.
Yet sports were, in the researchers' own words, "surprisingly absent" from the list of markets where arbitrageurs actually extracted significant profit. Politics dominated the leaderboard despite having fewer raw opportunities.
Why the disconnect? The researchers suggest sports opportunities tend to be smaller in individual size (1–5% margins vs larger political event dislocations) and more fragmented across thousands of individual matches. Automated arbitrage bots focused on the larger, more concentrated political markets where a single trade could capture more value.
This creates an interesting dynamic: the category with the most mispricing is also the most underexploited. For traders willing to systematically scan sports markets, the opportunity set is large and relatively uncontested.
What This Means for Prediction Market Trading
The paper validates several things that traders in this space have suspected but couldn't prove empirically.
Polymarket sports prices are measurably inefficient
This isn't speculation — it's demonstrated across 86 million data points. When compared against sharper pricing sources (traditional bookmakers with decades of odds-setting expertise), prediction market prices for sports events show consistent and exploitable deviations.
Buying NO is the dominant strategy
Of the 40 million dollars extracted, 17.3 million came from buying NO positions across market conditions. When a market overprices an outcome (YES is too expensive), the money is in buying the other side. This aligns with what experienced prediction market traders have observed — retail participants tend to overpay for the outcomes they want to happen, creating systematic overpricing of popular favourites.
The execution window is forgiving
Unlike high-frequency arbitrage in traditional finance where opportunities close in microseconds, the researchers found that 75% of arbitrage bids on Polymarket executed within approximately one hour. You don't need a co-located server and nanosecond execution — you need good data and timely alerts.
The margins are substantial
The median profit per mispriced condition was approximately 60 cents per dollar risked. That's not a few basis points of edge. That's a 60% return on correctly identified mispricing. Even accounting for partial fills and execution risk, the margins dwarf anything available in traditional financial arbitrage.
How WagerBase Fits In
WagerBase's MarketGap system does precisely what this paper describes as the underexploited opportunity: systematic, real-time scanning of sports prediction markets against traditional bookmaker pricing.
We compare bet365's odds (the sharp, vig-adjusted prices from one of the world's largest bookmakers) against Polymarket's prediction market prices across football, NBA, and tennis. When these two pricing mechanisms disagree — when there's a "gap" — we surface it as a trading signal.
The paper's findings map directly onto our data. Across 300+ settled MarketGap trades, we observe the same patterns the researchers identified: sports markets are consistently mispriced, the edge scales with the size of the disagreement, and the profitable side is typically aligned with the sharper bookmaker price.
Our MarketGap scanner tracks these gaps in real-time and alerts users via Telegram when new opportunities appear. The combination of the paper's empirical validation and our live tracking data paints a clear picture: prediction market sports pricing has measurable, systematic inefficiencies that persist despite the $40 million already extracted from the platform.
The question isn't whether the opportunity exists. The paper proves it does. The question is who's positioned to capture it as Polymarket's sports markets continue to grow.
Track live MarketGap opportunities and receive real-time alerts at wagerbase.io/odds. Follow us on X @Wagerbaseio.
Paper citation: Saguillo, O., Ghafouri, V., Kiffer, L., & Suarez-Tangil, G. (2025). Unravelling the Probabilistic Forest: Arbitrage in Prediction Markets. arXiv:2508.03474. Published in Proceedings of the 7th Conference on Advances in Financial Technologies (AFT 2025).