How Many Polymarket Traders Are Profitable?

No reliable public tally exists, and Convexly does not publish one. Treat any quoted percentage that arrives without a method as marketing. What the published record supports is narrower: in our own published top-50 cohort artifact, 12 of 50 wallets were net-negative on realized PnL, and a corrected platform-wide skill screen cleared 178 of 3,871 wallets. What you can always do is read one wallet honestly: the free Polymarket wallet analyzer works on any public 0x address, no signup.

Why no honest population tally exists

• The denominator is unknowable. Wallets are abandoned, recreated, and multiplied; one person can run several, and a dormant wallet with two losing trades from 2024 counts the same as an active account. Any “percentage of traders” inherits that noise.
• The numerator needs a window. Profitable since when? Open positions sit unresolved, and a wallet profitable over its lifetime can be deep underwater over the trailing year. A tally without a stated window is not checkable.
• Profit and skill are different counts. At retail sample sizes, a profitable record and a lucky record usually cannot be told apart from the resolved data; the readable question is whether entries beat the prices paid, with an interval.

What the published record supports

Convexly publishes a top-50 Polymarket cohort ranked by Edge Score, refreshed daily from a frozen 8,656-wallet reference cohort. In the published cohort artifact (2026-06-09), 38 of 50 wallets were positive and 12 of 50 net-negative on realized PnL by the artifact's sign column; the rendered leaderboard page shows the positive-PnL subset. Dollar figures are withheld because two of our own pipelines disagree on realized PnL for some wallets, and we do not publish numbers our own pipelines dispute. Even at the top of a skill-ranked board, profitability is not the default.

On the skill side, two published reads. A platform-wide screen of 3,871 testable wallets cleared 178 (about 4.6%) at a Benjamini-Hochberg false-discovery-rate threshold of q=0.10. And on 2026-06-09 we ran our skill read on our own published top-50 cohort; 35 of 50 had enough resolved positions to test; exactly one interval cleared zero on the positive side, roughly what chance predicts across 35 tests at a 2.5% threshold (about 0.9 false positives expected), and it did not survive a multiple-comparisons correction; none of the 50 are in the corrected cleared set. The full per-wallet table is at the top-50 skill scan. The careful phrasing matters: that is no corrected evidence of skill on those records, not evidence of no skill.

The question you can actually answer

“How many traders profit” is unanswerable from public data at the population level, but “does this specific record hold up” is answerable in about 30 seconds. The free check reads any public wallet's resolved record and returns calibration, sizing, concentration, and the realized entry edge with a bootstrap 95% interval against zero, with a 30-resolved-position floor below which no verdict renders. Who tends to end up on each side of the ledger is covered at who wins and who loses in prediction markets.

Diagnostics on public on-chain records, not investment advice; a past read is not a forecast.