Polymarket win rate, and why the profile number is always too high
If you have ever looked at a trader with an implausible win rate and wondered how anyone is that good, this is usually the answer. They are not.
What the number actually counts
A win rate is wins divided by resolved positions. The trap is in the denominator. On Polymarket a position only reliably enters that denominator when a settlement event exists, and losing positions frequently never generate one.
The reason is purely economic. Redeeming a winner collects a dollar a share. Redeeming a loser collects nothing and still costs gas, so nobody bothers. The tokens sit in the wallet, no settlement is written, and the loss is absent from anything computed off settlements.
So a published win rate is best read as an upper bound. The real figure is that number or lower, never higher.
How large the gap gets
On 11 August 2026 we rebuilt the five wallets ranked highest by weekly profit.
| Wallet (rank on 11 Aug) | Win rate from settlements | Real win rate |
|---|---|---|
| quavoo (#1) | 95.8% | 44.1% |
| ferrariChampions2026 | 66.4% | 49.8% |
| Djdjdjekekek | 91.0% | 48.6% |
| HOG993 | 78.1% | 65.7% |
| pleaseplease123 | 99.5% | 47.4% |
Not rounding differences. Different conclusions about the same five people in the same week, and four of them land within three points of a coin flip.
Notice that the correction is not uniform. HOG993 loses twelve points and stays clearly above even. The trader on the bottom row loses fifty-two. The size of the gap is itself information: it tells you how much of a record was never written down.
The counterintuitive part
A low win rate is not evidence of a bad trader, and a high one is not evidence of a good one. The wallet above loses more markets than it wins and is up substantially, because it sizes heavily when it is right. Payoff asymmetry does the work, not accuracy.
Which means a trader with a real 40% win rate and disciplined sizing can comfortably out-earn one at 70% who bets flat. Win rate on its own tells you almost nothing without position sizes next to it.
Computing the real number
- Pull the wallet's complete activity history, walking backwards by timestamp rather than by offset so nothing is silently truncated.
- Replay it into one position per market outcome. Splits and merges have to be replayed too, or the legs they mint look like sales of something never bought.
- For each market, look up how it actually resolved.
- Settle whatever is left over against that resolution. A unit carried through a winning resolution is worth a dollar and a unit carried through a losing one is worth nothing, redeemed or not.
Then divide. That denominator is the honest one.
A shortcut worth avoiding: treating "still held, never redeemed" as proof of a loss. It is right most of the time, which is what makes it dangerous. It cannot separate a worthless position from one whose market is still open, and it turns an unclaimed winner into a loss.
If you are grading wallets at scale
The bias is not random, which is the part that hurts. Wallets that lose most often accumulate the most unredeemed junk, so the correction is largest exactly on the wallets you are trying to identify as bad. A scan across thousands of wallets will systematically flatter the worst ones unless every wallet is diffed individually.
Common questions
- Why is my Polymarket win rate different from what I expected?
- Most likely because positions that resolved against you were never redeemed, so they are not counted as losses anywhere. Redeeming a worthless position costs gas and pays nothing, so almost nobody does it, and the loss leaves no settlement event.
- Is a high win rate on Polymarket a good sign?
- On its own, no. It is an upper bound rather than a measurement, and it says nothing about position sizing. A trader winning 40% of the time with disciplined sizing can out-earn one winning 70% betting flat.
- How do I calculate a real Polymarket win rate?
- Rebuild the full trade history, compare it against what the wallet currently holds, and treat anything missing from the live holdings that was never sold or redeemed as a loss at zero. Then divide wins by the corrected total.
- Does an unredeemed losing position still cost me anything?
- The money is already gone, it was spent when you bought. What it costs you afterwards is accuracy: it sits in your wallet looking like an open position, so your own dashboard overstates both your holdings and your record.