Polymarket win rate: why 100% means nothing without a sample size
A win rate is the most quoted number about a Polymarket trader and the easiest to misread. Read it with the number of markets behind it, through its confidence interval — and know what it leaves out.
What counts as a win
PMWallets scores each market a trader was in. A market counts once it has resolved by the end of the window and the trader traded or held it during the window: a win if the trader's profit on it is above one cent, a loss if it is below minus one cent, and a breakeven otherwise. Breakevens are left out, and the win rate is wins ÷ (wins + losses). Profit per market includes everything — fills, fees, splits, merges, redemptions and transfers — not just whether the last trade was on the winning side.
Why 100% can mean almost nothing
Three wins out of three is a 100% win rate, and a coin would manage it one time in eight. Seventy wins out of a hundred is a 70% win rate that luck rarely produces. Showing both as a bare percentage hides the one thing that tells them apart: how many markets stand behind it.
The confidence interval and its lower bound
PMWallets gives every win rate a Wilson 95% confidence interval and filters the board on its lower bound — a value the true win rate is unlikely to be below, given the sample. Three out of three has a lower bound of about 44%. Fourteen out of twenty, a 70% win rate, has about 48%. Seventy out of a hundred: about 60%. Seven hundred out of a thousand: about 67%. The same 70% means very different things, and the lower bound says how much.
What the interval does not fix
The interval treats markets as independent. Many are not: the markets of one multi-outcome event, the games of one tournament, or consecutive Up or Down windows on the same coin move together, so a run of related wins is less evidence than the same number of unrelated ones. Treat the interval as a floor on how cautious to be, not a guarantee.
Win rate is not profit
A trader who buys favourites at 95¢ can win 95% of markets and still lose money on the 5% they lose. One who buys long shots at 10¢ can win a third of the time and make a lot. Read the win rate next to realised profit, profit without the single best market, and the prices the trader typically pays.
Questions
What is a good win rate on Polymarket?
It depends on the prices a trader pays — buying at 90¢ needs a win rate above 90% just to break even. Compare the lower bound of the confidence interval, over many decided markets, with the trader's realised profit rather than the raw percentage.
What is a Wilson confidence interval?
A standard interval for a proportion that behaves well with small samples and rates near 0% or 100%. Its lower bound is a cautious estimate: given the wins and losses observed, the true rate is unlikely to be lower.
Why does PMWallets leave breakeven markets out of the win rate?
A market where a trader made or lost less than a cent — often dust left over from an earlier position — says nothing about skill. Counting it as a loss or a win would move the rate without any information.