Skill Leaderboard
Ranked Polymarket traders who made a trade in the last 180 days.
All-time grading and ranking methodology
Polymarket top traders, P&L, and grades in one table.
Each wallet's grade, volume, markets, realized P&L, and win rate side by side, so a big number is never read without its sample size.
Polymarket top traders
The table ranks by realized P&L, the money a trader banked. A skill tie-breaker across edge, track record, and consistency separates near-identical totals.
Trader P&L context
Realized P&L is about 95% of the grade and sits next to win rate, volume, and markets traded. Guardrails screen out wallets holding large unrealized gains, or only a handful of resolved markets, so a paper number cannot buy a top grade.
Best traders ranked
Wallets graded B or better make the board, each with verified net-positive realized P&L and a real resolved-market record behind the number. C, D, and F wallets are graded on their profiles but not ranked here.
How the leaderboard is scored.
Realized P&L, the trader's banked winnings (all-time P&L minus value still tied up in open positions), is about 95% of the grade. The remaining roughly 5% is a skill tie-breaker from forecasting edge, a verified track record, win rate, and risk-adjusted drawdown, and it only separates near-identical realized P&L; small samples are held back there with Wilson lower bounds and shrinkage. Any grade above C needs verified net-positive realized P&L, and the top grades also need a real resolved-market record and a survivable worst drawdown, so a wallet holding large unrealized gains, or only a handful of resolved markets, cannot reach the public S, A, or B board.
Banked winnings: all-time P&L minus the value still tied up in open positions. It is about 95% of the grade. The board ranks who banked the most money, not who holds the biggest paper gain.
Whether the trader's implied probabilities beat the market price. It is part of the roughly 5% skill tie-breaker between near-identical realized P&L. Small samples are shrunk toward the cohort, and bet size does not count.
How many markets the trader resolved, on a log scale. The top grades need a real resolved-market history, so a wallet with a handful of settled markets is screened out however big its total.
Resolved-market win rate held to its Wilson lower bound, so a short hot streak does not count like a long record. It is a tie-breaker.
Shallower worst-ever drawdowns and a smoother equity curve rank higher, alongside a downside-aware risk-adjusted return. Any grade above C needs verified net-positive realized P&L. The top grades also need a real resolved-market track record and a survivable worst drawdown, so a net-losing, unproven, or paper-gain wallet cannot reach the public S, A, or B board.
A thin record does not reach the top.
Realized P&L ranks the board. The skill tie-breaker stays conservative until enough markets resolve: win rate is held to its Wilson lower bound and return is shrunk toward the cohort. The top grades also need a real resolved-market record, so a wallet that banked a big number on a handful of settled markets does not sit at the top.
- The skill tie-breaker tightens as more markets resolve.
- Thin samples stay anchored near the cohort until the edge repeats.
- Top grades require verified realized P&L and a real resolved-market history.
Three rules decide the public table.
- Grades are relative to the whole graded cohort and recompute about every hour, so they shift as other traders improve or fade.
- Only grades S, A, and B with a known trade in the last 180 days make the Skill Leaderboard.
- A wallet with fewer than 5 markets has no grade, and C, D, and F wallets are not on the board.
Win rate rewards buying favorites. Edge checks the price.
A 99¢ contract wins about 99 times in 100, so a wallet that only buys near-certain outcomes tops a win-rate sort without showing any skill. Edge is how often the side a wallet bought won, minus the average price it paid. Win 70% of markets bought at 55¢ on average and the edge is +15 percentage points; win every 99¢ contract and it is about +1.
- Counted over settled markets of $20 or more with a known entry price.
- Shown and ranked at the low end of a 95% range, so a short lucky run cannot top the column.
- A wallet needs 30 such markets, and the side it bought must win them within 10 points of its overall win rate, or it gets no edge rank.
# how a grade is computed
# cohort (graded population)
markets_traded >= 5
# S/A also need volume >= $50K / $20K and positive P&L + edge
# above C needs positive realized P&L and 10+ resolved markets
pnl = 0.5*pct(realized_pnl) + 0.5*pct(realized_pnl_90d)
skill_index = 0.95*pnl + 0.05*skill_blend(edge, edge_90d, track, win, drawdown, smoothness, sortino)
skill_pct = cohort_percentile(skill_index)
grade = bands(skill_pct), then floors and idle caps
# _90d: each day weighted by a 90-day half-life
# idle 30+ days: capped at B; idle 120+ days: capped at C
# no 90-day history yet: the last grade is kept, floors and caps still apply
# missing signals excluded, not defaulted to perfect
# realized P&L is ~95% of the grade; floors screen unproven and paper-gain wallets off the top
# the whole cohort is regraded about every hour
traders ranked (grade B+)