A Bitquery audit found that betting on every favorite in Polymarket's 2026 US primaries would have cost traders 4 percent, even as those picks won 87 percent of the time. The results call prediction market pricing into question.
Traders who put $1 on every favorite in Polymarket's 2026 US primary races would have ended up with less cash than they started with. Bitquery's audit shows that, even as favorites won the vast majority of races, the strategy bled money. The numbers: 273 scored Senate, House, and governor primaries, 238 favorites won, but the approach lost four cents per dollar wagered. The audit exposes a stubborn gap between picking winners and actually making money. Forecasting right doesn't always mean the trade pays off.
Bitquery's October 9 analysis tracked 273 US primary contests on Polymarket, a blockchain prediction market. The favorite-defined as the candidate with the highest reference price-came out ahead in 238 races. Yet, the $1-per-favorite method still lost money. Reference prices were set by averaging trades in the 24 hours before noon UTC on voting day, or by using the last trade within 30 days if recent activity was missing. Runoff races used the runoff date. These reference prices don't always match real buy quotes at the cutoff, and the audit doesn't fully adjust for spreads, slippage, or fees.
Polymarket reported over $6 billion in trading volume in the first half of 2025, highlighting its rapid growth as a blockchain-based prediction market.
Nine markets were left out due to early settlement, missing prices, or incomplete records. The audit only looked at Polygon-based Polymarket trades, skipping the US app and other venues like Kalshi. Polymarket's current fee page lists a politics fee range from 0 to 1 percent, but the audit doesn't specify what fees actually hit the trades in the sample.
The headline win rate came from the strongest favorites. Candidates priced at 90 cents or higher won 177 out of 182 races. Among favorites priced between 50 and 90 cents, only 71 percent won, even though the average price paid was 77 cents. Each outcome share pays $1 if correct and nothing if wrong. Buying expensive favorites leaves little room for profit, and a single loss wipes out the purchase cost. Enough losses among mid-priced favorites can erase gains from the high-probability picks.
Even when the favorite is likely to win, the contract price can already bake in that probability so tightly that there's no edge left for the bettor. The difference between forecasting accuracy and profitable trading often catches new users off guard. As reported earlier, market structure and pricing can hide risks for traders who assume high win rates mean easy money.
Polymarket contracts typically trade between $0.01 and $0.99, directly reflecting the market-implied probability of an event. Fees are dynamic and tend to be higher for contracts near 50/50 outcomes, while maker orders incur no commission and taker trades are charged between 0.01% and 1.80% depending on category and contract price.
Bitquery's methodology leans on reference prices that can be stale or unrepresentative, especially in thin or illiquid markets. The audit skips over some trading costs, including spreads and slippage, which can eat into returns. The study also points out that general elections draw more money and polling attention than local primaries, so these patterns may not show up in bigger, more liquid markets. For November's prediction markets, the split between forecasting and pricing remains a live issue for traders and market designers.
Bitquery covered only Polygon-based Polymarket trades for US primaries in all 50 states, leaving out other platforms and the Polymarket US app. The audit's findings stick to the specific sample and method described, and don't promise similar results in future or different market conditions.
Polymarket's fee structure for politics markets now ranges from 0 to 1 percent, but the audit doesn't clarify what historical fee rates applied to the trades analyzed. The lack of full adjustment for fees, spreads, and slippage means actual realized returns could differ from the hypothetical results shown.
Bitquery's audit found that, as of October 9, 2026, backing the favorite in every scored US primary on Polymarket would have lost 4 percent overall, even though favorites won 87 percent of the time. The strongest favorites-those priced at 90 cents or more-won 177 out of 182 races, while mid-priced favorites (50 to 90 cents) won 71 percent of the time at an average price of 77 cents. The audit covered 273 races, using Polygon trade data and reference prices based on trading activity in the 24 hours before noon UTC on voting day or the last trade within 30 days. Nine markets were excluded due to early settlement, missing prices, or incomplete records.
Prediction markets like Polymarket let users buy and sell outcome shares that pay $1 if the predicted event happens and nothing if it doesn't. The price of each share reflects the market's consensus probability, but trading costs, liquidity, and market structure can create a gap between forecasting accuracy and actual profitability. Even a strategy that nails the most likely winner most of the time can lose money if the contract price leaves no margin for error or if trading costs are underestimated. For users, knowing the mechanics of pricing, fees, and liquidity is the only way to avoid costly surprises.
