The Biggest Prediction Market Upsets: Heavy Favorites That Lost

By Arthur Angelo, PredictionScout Research · Updated Oct 11, 2026, 9:48 AM ET

Key findings

• Heavy favorites rarely lose, but it happens. Of 15,977 questions where the favorite cost 90¢ or more one day before the outcome was clear, 773 lost (4.8%). At 95¢ or more, 328 of 12,432 lost (2.6%).
• Sports vs everything else. Outside sports, 535 of 13,124 90¢ favorites lost (4.1%); in our sports sample, 238 of 2,853 (8.3%).
• Kalshi: 443 of 7,171 90¢ favorites lost. Polymarket: 330 of 8,806.
• Biggest non-sports miss in our sample: “Will the price of Ethereum be above $2,200 on August 20?” on Polymarket, where No traded at 99.9¢ a day before the outcome was clear and lost (Aug 20, 2026).

Where 90¢ favorites lost most often

The biggest upsets outside sports

Sorted by how expensive the losing favorite was one day before the outcome became clear. “Decided” is the date the market settled on its final answer. Click a market name to see it on the exchange.

The biggest sports upsets in our sample

From a random sample of sports events (see methodology), so this is not a complete list of every sports upset.

Download the data

Every upset (CSV)

Every question where the favorite was priced 85¢ or more one day before the outcome was clear and lost: market, exchange, category, favorite, prices 1 day and 1 week out, the winner, decision date, volume and link. Free to reuse under CC BY 4.0.

Methodology

• Sample. The settled markets in our accuracy scorecard: Kalshi markets with at least 50,000 contracts traded and Polymarket markets with at least $50,000 in volume, closed since January 1, 2025. Sports and crypto events are a fixed random sample; every other category is complete above those floors.
• Upset. The favorite (highest-priced outcome, or the side above 50¢ in a yes/no market) cost 85¢ or more exactly 24 hours before the outcome became clear, and lost. “Outcome became clear” is the first hour after which the winner’s price stayed at 99¢ or more until close, or the close if that never happened.
• Prices. Kalshi: midpoint of the hourly bid and ask when the spread is 10¢ or less, otherwise the last trade. Polymarket: hourly price history from its public API.
• Results are the exchanges’ own settlements. A few markets settle on technical wording (an exact phrase, a data source, a deadline) that can surprise traders; we have not re-checked each settlement against outside sources. Explanations appear only where we have checked them against a cited source.
• Expect upsets. If 90¢ prices are accurate, about one in ten should lose. A list like this shows that markets can be wrong, not that they are badly calibrated; for that, see the price tables in the accuracy scorecard.
• Updates. Weekly, together with the accuracy scorecard. Last run: Oct 11, 2026, 9:48 AM ET.

Data sources: prices and volumes come from the public APIs of Kalshi and Polymarket. All trademarks belong to their owners. PredictionScout is independent and not affiliated with any exchange. Nothing here is investment, betting, or legal advice.

Related research: Accuracy scorecard · Unusual price moves before results · All research