Polymarket favorites would have lost 4%, audit finds
A hypothetical technique of backing the favourite on prediction market Polymarket in each scored 2026 US main would have misplaced 4%, despite the fact that these candidates gained 87% of the time, in response to blockchain information supplier Bitquery’s scorecard dated Oct. 9. The consequence separates two questions for election bettors: who’s more than likely to win, and whether or not their contract is definitely worth the worth.
Bitquery discovered that the candidate with the best research reference worth gained 238 of 273 Senate, Home and governor primaries. However its hypothetical technique of placing $1 on every favourite on the research’s reference worth misplaced 4 cents per greenback.
The strongest favorites did a lot of the work behind the headline accuracy fee. Candidates priced at 90 cents or extra gained 177 of 182 races. Amongst favorites priced between 50 and 90 cents, nevertheless, 71% gained regardless of a mean worth of 77 cents.
An final result share pays $1 when it wins and nothing when it loses. Shopping for an costly favourite due to this fact leaves little revenue on a successful share, whereas a dropping share wipes out its buy price. Sufficient losses can outweigh many right calls.
Placing the identical greenback quantity into each race additionally buys completely different numbers of shares. A less expensive winner produces a bigger payout for that stake than an costly winner. Counting right predictions treats each race equally; calculating returns should account for these completely different payouts.
Reference costs restrict the betting conclusion
Bitquery averaged trades throughout the 24 hours earlier than 12:00 UTC on voting day. When a candidate had no commerce in that window, it used the final commerce inside the previous 30 days. For races that went to a runoff, it used the runoff date.
These averages, and probably stale fallback trades, are reference costs reasonably than assured purchase quotes on the cutoff. The revealed methodology doesn’t specify a full adjustment for spreads, slippage or charges. Polymarket’s current fee page lists a politics price vary of 0% to 1%, however doesn’t set up the costs on the historic trades on this pattern.
The research excluded 9 markets as a result of some settled earlier than voting, some lacked usable costs and one lacked a settlement file. It coated primaries within the 50 states utilizing Polygon commerce information, leaving out Polymarket’s US app and different venues equivalent to Kalshi.
For November’s prediction markets, the excellence between forecasting and pricing stays helpful. The measured main file itself might not carry over: Bitquery cautions that normal elections entice more cash and polling than small, native contests.
