Why PnL Lies About Skill
The most-followed Polymarket wallets are almost always the biggest ones. That feels right — until you notice that the biggest wallets are also the ones taking the most risk, on the most markets, with the most money. Profit and loss is a measure of stakes, not skill.
The size problem
Imagine two traders. One bets $500,000 across a year and finishes $40,000 up. The other bets $5,000 and finishes $2,000 up. The first shows a bigger number. The second is the better forecaster — they turned a far higher fraction of their exposure into profit, and they did it without needing a whale's bankroll.
A leaderboard sorted by raw profit is, in practice, a leaderboard of bankroll. It tells you who is rich, not who is right.
What skill actually looks like
Skill is the ability to pay a price that is lower than the eventual outcome, repeatedly, across many independent markets. If you buy "Yes" at $0.41 and it resolves to $1.00, you captured $0.59 of edge per dollar — regardless of whether you staked ten dollars or ten thousand.
That per-dollar, outcome-relative quantity is what we score. Aggregated and weighted by stake, it becomes value-weighted excess hit rate (EHR_vw) — the metric from Akey, Grégoire, Harvie & Martineau (2026) that anchors our leaderboard.
“Winners are makers. The edge shows up in the prices people are willing to accept, not in the size of their bankroll.
”
Three properties that matter
- Outcome-relative. Every trade is scored against where the market actually settled, so a cheap winning bet beats an expensive one.
- Stake-weighted, not stake-driven. Bigger conviction counts for more, but raw bankroll never substitutes for being right.
- Bounded by sample. A wallet with five resolved trades is not a genius. We publish every score with a confidence interval and a trade count.
Reading a score honestly
Here is the number that started the project — the share of Polymarket profit captured by the top sliver of wallets:
Concentration like that is exactly why a naive PnL board is misleading: a handful of large, lucky-or-skilled wallets dominate the totals and bury the forecasters worth following.
| Wallet | Total PnL | EHR_vw | Verdict |
|---|---|---|---|
| Whale A | $210,000 | +0.004 | Big, not sharp |
| Quiet B | $9,400 | +0.087 | Sharp, not big |
The whale wins the PnL column. The quiet wallet wins the only column that predicts the next trade.
What we do with it
Every wallet on the leaderboard is scored on resolved trades only, normalized to the buy side, clustered to defeat sybil splitting, and published with its interval and sample size. No bare numbers, no survivor-only boards, no claims we can't show the math for.
You can see the result ranked this way on the top 100 Polymarket wallets board — every entry with its confidence bound, sample count, and receipts. For the market structure behind it, read the 1% who own 76% of Polymarket.
That is the whole thesis: follow the edge, not the bankroll.
FAQ
Why is raw PnL a bad way to rank Polymarket traders? PnL rewards size and luck: the biggest wallets take the most risk on the most markets, so profit measures stakes, not skill. A wallet can be hugely profitable and statistically indistinguishable from a coin-flipper.
What should you use instead of PnL? Value-weighted excess hit rate (EHR_vw) — how much better a wallet buys than the market price implied, weighted by position size, with a 95% confidence bound and a resolved-trade count so luck is visible.