A hypothetical strategy of backing Polymarket's favorite candidate in every 2024 US primary race would have lost 4%, despite those candidates winning 87% of the time, according to an analysis by blockchain data provider Bitquery dated October 9.
The finding reveals a distinction between predictive accuracy and betting profitability. Bitquery identified the candidate with the highest reference price as the favorite in 273 Senate, House, and governor primaries. Those favorites won 238 races, but an equal-dollar betting strategy on all favorites at the study's reference price lost four cents per dollar wagered.
Price Matters
The strongest favorites drove much of the accuracy rate. Candidates priced at 90 cents or higher won 177 of 182 races. Among favorites priced between 50 and 90 cents, however, 71% won despite an average price of 77 cents.
The mismatch between accuracy and returns reflects how prediction market contracts work. An outcome share pays $1 when it wins and nothing when it loses. Expensive favorites leave little profit on winning shares, while losses eliminate the full stake. When applied across many races with equal dollar amounts, enough losses can outweigh numerous correct predictions.
Equal-dollar betting also purchases different quantities of shares. A cheaper winner generates a larger payout for that stake than an expensive winner, creating unequal returns even when predictions are correct.
Study Methodology
Bitquery calculated reference prices by averaging trades during the 24 hours before voting day. When no trade occurred in that window, it used the last available trade within the preceding 30 days. For runoff races, it used the runoff date.
The analysis excluded nine markets due to early settlements, unavailable prices, or missing settlement records. It covered primaries across the 50 states using Polygon trade data, excluding Polymarket's US app and other venues.
The study does not specify full adjustments for spreads, slippage, or fees on historical trades. The distinction between forecasting accuracy and pricing remains relevant for future prediction markets, though the primary record may not carry over to general elections, which typically attract more money and polling activity than local contests.
