Research
We Analyzed 88,000 Prediction Market Contracts. Here's Where the Edge Is.
Most prediction market traders lose money. But a structural pricing bias — documented across decades of academic research — creates a measurable statistical edge for those who know where to look.
Prediction markets are having a moment. Platforms like Kalshi and Polymarket have crossed billions in trading volume, attracting everyone from institutional traders to curious newcomers betting on everything from Federal Reserve rate decisions to the weather.
The appeal is intuitive: if you think you know better than the market, you can profit from it. But the data tells a different story. The vast majority of retail traders on prediction markets lose money — just like in sports betting, options trading, and every other speculative market.
Except there's a pattern. A structural anomaly hiding in the price data that has been documented across horse racing, sports betting, and now prediction markets. It's called the Favourite-Longshot Bias, and it creates a measurable edge for traders who understand it.
The Favourite-Longshot Bias
The Favourite-Longshot Bias (FLB) is one of the most well-documented anomalies in betting markets. The core finding is simple: longshots are systematically overpriced, and favourites are systematically underpriced.
In horse racing, this means that horses with low odds of winning (longshots) tend to win even less often than their odds suggest, while heavy favourites win more often than their odds imply. The pattern was first documented by Griffith in 1949 and has been replicated across hundreds of studies since.
The mechanism is psychological. Humans systematically overestimate the probability of unlikely events — the lottery ticket effect. On prediction markets, this manifests as traders overpaying for low-probability "YES" outcomes (the longshot), which means the "NO" side (the favourite) is underpriced.
In 2026, Bürgi, Deng & Whelan published the first comprehensive analysis of the FLB on Kalshi, analyzing 88,570 settled contracts across all categories. Their findings confirmed the bias exists — and it's tradeable.
Our Analysis: 88,570 Contracts, 120 Days
We built on this research with our own debiased historical analysis. Over a 120-day period, we analyzed every Kalshi contract that:
- Had a NO-side price between 82¢ and 92¢ (meaning the market implied an 82–92% probability of the event occurring)
- Had at least 100 contracts traded in the last 24 hours
- Was not in excluded categories (Mentions, Climate/Weather)
- Had a maker spread of 4¢ or less
The results were striking. Contracts in the 84–92¢ NO price range settled in favour of the NO side 92.8% of the time — significantly higher than the 88.3% implied by the average market price. This represents an implied statistical bias of +4.5 percentage points.
88,570
Contracts analyzed
92.8%
Actual NO win rate
88.3%
Implied by price
+4.5%
Statistical edge
The Data
Let's break down the data by price bucket. Each bucket represents a 2-cent range of NO-side entry prices, and we measured the actual win rate against the market-implied probability.
Every single bucket outperformed the market-implied probability (the red dashed line at 88.3%). The 86–88¢ and 90–92¢ buckets were particularly strong, with win rates above 95%.
The bias also holds across categories, though some are stronger than others:
Financials and Crypto showed the strongest edge, likely due to higher liquidity and more systematic pricing. We excluded Mentions and Climate/Weather categories from the strategy after finding they had negative implied bias — the market actually underpriced longshots in these categories.
Cumulative Performance
We backtested a systematic strategy: buy NO-side contracts at 84–92¢ using maker orders only, hold to settlement, size positions at 1% of capital per trade, and diversify across correlation groups.
Over 120 days, the strategy generated +$421.60 in cumulative PnL on a $1,000 starting balance, with relatively smooth equity growth. The brief drawdown around day 40 coincided with a period of high volatility in the crypto markets, but the strategy recovered within two weeks.
Why It Works
Three structural factors drive the Favourite-Longshot Bias on prediction markets:
1. Human psychology. People systematically overestimate the probability of unlikely events. On prediction markets, this means traders overpay for low-probability YES outcomes — the equivalent of buying lottery tickets. This inflates YES prices and depresses NO prices, creating the edge.
2. The lottery ticket effect. A contract priced at 10¢ (90% chance of NO) offers a potential 10x return if YES happens. This asymmetric payoff attracts disproportionate speculative interest on the YES side, even when the probability doesn't justify it.
3. Market structure. Taker fees on Kalshi (7%) are significantly higher than maker fees (1.75%). Most retail traders are takers, which means their breakeven threshold is higher. Makers — who post limit orders at the bid — capture the spread and pay lower fees, amplifying the edge.
How to Trade It
If you want to trade the FLB systematically, here are the principles that our research supports:
Buy NO on high-probability contracts. Focus on contracts where the NO-side price is between 84¢ and 92¢. Below 84¢, the edge is thinner. Above 92¢, win rates actually decline (the "certainty trap").
Use maker orders only. Post limit orders at the NO bid price. This gives you the 1.75% maker fee instead of the 7% taker fee — a massive difference that turns marginal trades into profitable ones.
Diversify across categories and settlement dates. Don't concentrate in one market or one settlement time. Spread your positions across Financials, Crypto, Economics, and other categories to reduce correlation risk.
Monitor your win rate over 50+ trades. Any strategy needs sufficient sample size to evaluate. Track your rolling win rate and compare it to your average entry price. If your win rate is consistently above your average entry, you have a positive implied bias — keep going.
Or let Ledion find these opportunities for you. We scan every Kalshi market every 5 minutes, rank opportunities by our composite scoring algorithm, and surface the highest-edge contracts directly to your dashboard.
Track these opportunities in real-time
Ledion scans thousands of prediction market contracts and surfaces the ones with the highest statistical edge. Free to start, no credit card required.
Methodology
Data sourced from Kalshi's public API covering all settled contracts from November 2025 through March 2026 (120 days). 88,570 contracts met the filtering criteria. Backtest uses debiased methodology: each contract is evaluated at the price available at the time of the scan, not with hindsight. Maker fees (1.75%) are deducted from all trades. Past performance does not guarantee future results. Based on Bürgi, Deng & Whelan (2026), "The Favourite-Longshot Bias in Prediction Markets: Evidence from Kalshi."