A hypothetical strategy of betting $1 on the frontrunner in every scored 2026 U.S. primary election market on Polymarket would have resulted in a 4% loss, despite those candidates winning 87% of their races, according to an October 9 audit by blockchain data provider Bitquery.
Key Takeaways
- High Win Rate, Negative Return: Favorites won 238 out of 273 Senate, House, and gubernatorial primaries, yet a $1 stake on every favorite lost four cents per dollar.
- Extreme Odds Drove Accuracy: Candidates priced at 90 cents or higher won 177 of 182 races, while candidates priced between 50 and 90 cents won 71% of the time despite an average price tag of 77 cents.
- On-Chain Polygon Scope: The Bitquery study analyzed Polygon transaction data across all 50 states, excluding nine markets due to early settlement or missing price records.
Asymmetric Risk Wipes Out High Accuracy Rates
The disparity between a high win frequency and negative investment returns stems from the mathematical structure of prediction market contracts. On platforms like Polymarket, a winning share pays out $1.00, whereas a losing share drops to zero. Consequently, purchasing an expensive favorite at 90 cents yields a profit of only 10 cents per winning share, while a single loss wipes out the entire 90-cent investment.
In the middle pricing tier, Bitquery found that candidates priced between 50 cents and 90 cents carried an average price of 77 cents but won only 71% of the time. This pricing gap meant that losses in mid-tier contests dragged down overall strategy returns, exceeding the small profits accumulated from heavy favorites. Similar risk dynamics and outcome distributions have been highlighted across crypto research, such as how data studies analyze trading returns.
Methodology and Market Limitations
Bitquery calculated reference prices by averaging trade execution data during the 24-hour window leading up to 12:00 UTC on voting day. When a candidate had no trades within that window, the study referenced the last executed trade within the preceding 30 days, or used the runoff election date where applicable.
The audit evaluated primary contests on Polygon and did not include Polymarket’s U.S. app or alternative prediction venues such as Kalshi. Furthermore, the methodology did not fully adjust for order book spreads, slippage, or trading fees, which Polymarket currently lists between 0% and 1% for political contracts. Bitquery noted that general elections attract significantly higher liquidity and polling volume than small local contests, which may alter pricing efficiency in upcoming November markets.
Why It Matters
This audit highlights a critical distinction in decentralized prediction markets: high forecasting accuracy does not automatically guarantee a profitable trading strategy. As event contracts gain popularity among retail participants, mistaking high probability for positive expected value remains a significant risk. For crypto traders evaluating market efficiency and capital allocation—similar to how institutional reports assess capital requirements and trading efficiency—understanding asymmetric contract pricing is essential before deploying funds into prediction pools.
