Polymarket Prediction Accuracy at 87% Still Led to 4% Loss on Equal Betting

Polymarket saw an 87% accuracy rate in predicting US primary election outcomes, but assuming equal dollar bets on candidates each voting day led to a 4-cent loss per dollar. Blockchain data firm Bitquery analyzed 273 primary elections for the 2026 Senate, House, and gubernatorial races on October 9. The candidate with the highest Polymarket price on election morning won in 238 of those cases.
Prediction market contracts pay $1 per share if correct, nothing if wrong. Bitquery noted that win rate alone doesn't determine profitability since purchase price matters too. Among candidates trading above $0.90, 177 out of 182 won (97% accuracy). In the $0.50-$0.90 range, win rate was 71% with an average price of $0.77.
Bitquery stated that under equal dollar investment assumptions, there was a 4-cent loss per dollar overall. The analysis has limitations: Bitquery averaged prices from the previous 24 hours before noon election day (KST), using last available price if no trades occurred in that window or final election day for runoff races. This calculated price may not match actual execution prices due to spreads, slippage, and fees which Bitquery didn't fully account for.
The analysis covered Polymarket transactions verified on Polygon (Polygon), excluding other platforms like Kalshi, markets settling before voting or with insufficient pricing data, and 9 cases with missing settlement records. Bitquery cautioned that primary election results don't directly translate to November general elections due to higher funding and polling in those races.
This 4% loss was based on historical price calculations, not actual execution prices or transaction costs for individual traders.
Korean Source
This article is an English localization of a Korean-language crypto news report. Original headline: 폴리마켓 예비선거 적중률 87%에도 동일 금액 베팅은 4% 손실