The Polymarket leaderboard is a ranked, refreshable view of the top Polymarket traders, scored on a composite that blends Sharpe, ROI, edge-adjusted hit rate, and drawdown. Read it column by column — not row by row — and use sample size, category mix, and concentration as primary filters before composite score. Poly Syncer indexes 12,438 wallets and republishes the ranked list every 15 minutes; this guide explains what each column means, what filter recipes work for each goal, and the four pitfalls that catch new users.
Why the leaderboard exists at all
Polymarket is a public order book on an EVM-compatible chain. Every fill, every wallet, every position is observable. That is enormously useful and enormously noisy — you can see the trades of 12,438 wallets, but the median wallet is a roundtripping novice, the long tail is dominated by sub-30-trade samples, and the best wallets are scattered across categories that have wildly different liquidity and volatility profiles. A leaderboard is the compression layer: it ranks wallets so that "look at the top of the list" produces a useful starting set rather than a random sample.
The Poly Syncer leaderboard, served at /leaderboard, is one such ranked view. Its scoring rules are documented in detail at /methodology and our companion post on wallet scoring methodology. This guide is the user-facing manual: how to read what is on the screen.
The columns, one by one
The default view shows ten columns. Each one is doing a specific job; knowing what the job is keeps you from misreading the table.
| Column | What it answers | Useful range |
|---|---|---|
| Rank | Position in the composite | 1–100 is the working set |
| Composite | Single-number summary | +1.0 to +2.5 is "good" |
| Sharpe (30d) | Return per unit of volatility | 1.6–3.0 sustainable |
| ROI (30d / 90d) | Realized percentage return | Context-dependent |
| Edge HR | Win-rate minus avg implied prob | +0.04 to +0.12 healthy |
| Trades | Sample size | ≥100 preferred |
| Max DD | Worst drawdown in window | Below 15% ideal |
| Categories | Breadth flag | 2–6 is a sweet spot |
| Concentration | HHI on USDC volume | Below 0.4 |
| Last active | Recency | Within 72h preferred |
Rank vs composite: read both
Rank is ordinal; composite is cardinal. The wallet at rank 7 with composite +1.81 and the wallet at rank 12 with composite +1.79 are statistically indistinguishable. Treat the leaderboard as a set at the top — the top 25 or top 50 — not as an ordered ladder.
Sharpe ratio
The Sharpe ratio here is computed on trade-bucketed returns rather than time-bucketed returns, because Polymarket positions are discrete events rather than continuous prices. A 30-day Sharpe above 2.0 puts a wallet in roughly the top decile of the indexed cohort. Above 5.0 should make you suspicious of sample size or outlier dominance — check the winsorized Sharpe column on the wallet's profile page.
ROI windows
We publish 7-day, 30-day, and 90-day ROI. The 7-day number is mostly noise but useful to detect regime change: a wallet that has been excellent and is now cold. The 90-day number is the most predictive of the next 30-day window; in our internal data the rank correlation is roughly ρ = 0.41, versus ρ = 0.28 for 30-day-on-30-day.
Edge-adjusted hit rate
This is the column most users misread. A 78% win rate sounds amazing until you notice the wallet's average entry price is $0.81. The wallet is making roughly fair-priced bets at small edges; the win rate is a pricing artifact. Edge HR subtracts the average implied probability from the win rate, so a wallet at 78% wins on $0.81-priced contracts gets +0.03, not +0.78. Look for +0.04 to +0.12 across at least 100 trades.
Trades column — sample size
With 30 binary trades the standard error on win rate alone is about 9 percentage points; with 200 trades it is roughly 3.5 percentage points. A wallet with 32 trades and a 65% raw win rate could plausibly be a 50% wallet that got lucky. Sort by composite, then re-filter to trades ≥ 100: about 38% of the indexed cohort meets that bar.
Max drawdown
Drawdown answers the question "how bad did it get?" rather than "how good did it get?" A wallet with +60% ROI and −38% max DD is structurally different from a wallet with +35% ROI and −7% max DD — even if their Sharpes are similar — because copy-trading the first one means living through the drawdown in real time. Pair max DD with time-to-recovery, available on the profile page.
Categories
Polymarket has 25 active categories — politics, sports, crypto, earnings, weather, awards, macro, geopolitics, and more. A wallet that only ever trades a single category is a specialist; that is fine but it means you are exposed to that category's resolution cadence. A wallet that touches 12 categories is either a generalist with a real model or a noise generator. The 2–6 category band tends to correlate best with persistent edge in our data.
Concentration (HHI)
The Herfindahl-Hirschman index on USDC volume by market answers "how much of this wallet's return came from one position?" An HHI above 0.4 means a single market accounts for more than 40% of the wallet's volume. The wallet may still be excellent, but its score is fragile to one outcome.
Last active
Cold wallets — no fills in 5+ days — are dangerous to copy because you do not know if the operator has stopped, switched accounts, or is waiting for a specific catalyst. We surface a "stale" badge after 72 hours.
Sorting strategies for different goals
The default sort is composite descending. That is the right starting point for most users, but the column you sort by should match your goal.
Goal: smooth equity curve
Sort by Sharpe descending, then filter trades ≥ 200 and max DD ≤ 12%. You will get a shorter list (roughly 60–90 wallets across the indexed cohort) but the volatility you experience as a copier will be much lower. Pair with a tight risk-gate setup.
Goal: high upside, willing to ride drawdowns
Sort by 90-day ROI descending, ignore max DD, but require trades ≥ 100 and edge HR ≥ +0.05. You are accepting variance in exchange for tail upside. Position sizing on your end becomes the safety net — see our piece on Kelly sizing.
Goal: single-category specialist
Filter category = "Politics" (or whichever), then sort by composite. Specialists usually have lower trade counts (an election cycle is a finite resource) but higher edge HR than generalists in their domain. The categories reference covers liquidity and edge expectations per category.
Goal: long-tail value hunters
Sort by edge HR descending with average entry price filter ≤ $0.45. These are wallets making contrarian bets on under-priced contracts. They typically have lower raw win rates (40–55%) but the math is on their side. Sample size matters more here, not less — require trades ≥ 150.
Goal: high-frequency mirroring
Sort by trades descending, filter Sharpe ≥ 1.4 and Last-active within 24h. You want a wallet that puts plenty of fills into your engine so the copy ledger compounds quickly. The 600 ms p99 Elite execution lane is built for this; the MEV-protection guide explains why latency matters.
Filter recipes (saveable in your dashboard)
Each recipe below is a set of filter values you can paste into the leaderboard search. Save them under /dashboard/settings as named presets.
| Recipe | Filters | Typical result count |
|---|---|---|
| Conservative core | Sharpe ≥ 2.0, trades ≥ 200, max DD ≤ 10%, HHI ≤ 0.3 | ~45 wallets |
| Balanced | Composite ≥ +1.4, trades ≥ 100, last active ≤ 72h | ~140 wallets |
| Politics specialist | Category = Politics, edge HR ≥ +0.06, trades ≥ 80 | ~30 wallets |
| Sports specialist | Category = Sports, Sharpe ≥ 1.8, trades ≥ 150 | ~55 wallets |
| Crypto/macro | Categories include Crypto or Macro, edge HR ≥ +0.05 | ~70 wallets |
| Underdog hunters | Avg entry ≤ $0.40, win rate ≥ 45%, trades ≥ 150 | ~25 wallets |
Common pitfalls
Chasing recent winners
The single most common error is sorting by 7-day ROI and copying the top 5. Seven-day ROI has the lowest predictive power for next-window performance — it is dominated by single-event resolution noise. The wallet that just scored a 4x on a single politics resolution will look like a god for a week and then revert.
Ignoring sample size
A composite score on 38 trades is not the same as a composite score on 380 trades, even if the number reads identically. Always check the trades column. We shrink Sharpe toward the cohort median based on sample size (the formula is on /methodology) so the displayed Sharpe is already conservative for low-N wallets, but you should still apply a hard floor of trades ≥ 100 for any wallet you copy with non-trivial size.
Ignoring category mix
If you copy three wallets that are all 90% politics, you are not diversified. You are running a single concentrated politics book through three execution paths. Diversification at the wallet level only diversifies if the wallets trade different categories. A simple check: open each wallet's profile, look at the category histogram, and require ≥ 2 distinct dominant categories across your follow set of 5 wallets.
Treating composite as a guarantee
Composite is a screen, not a promise. The base rate of a top-50 wallet surviving the next 30 days inside the top-50 is roughly 41% in our backtests. That is much better than random (5%) but it is not 100%. The leaderboard tells you where to start your search, not where to stop your due diligence; the wallet evaluation checklist picks up where the leaderboard leaves off.
How the refresh cycle affects what you see
The leaderboard republishes every 15 minutes. Two implications:
- If a leader resolves a major position between refreshes, you may copy their entry using the previous-cycle scoring — that is intentional, the engine fires on the live trade, not on a stale composite.
- During major resolution waves (election nights, NBA finals), Spearman rank correlation between consecutive refreshes can drop from its ~0.92 baseline toward 0.85. We surface a banner when this happens; treat the rankings as wider error bars during that window.
Plan-level differences in what you see
The free leaderboard shows ranked wallets with composite, Sharpe, and ROI. Pro ($299) adds the full filter set, edge HR column, concentration, and saved presets. Elite ($499) adds the realtime watchlist push and the 600 ms p99 execution lane (Pro is roughly 1.5 s p99). Plan details and upgrade flow are on /dashboard/billing; the architectural reason for the latency tier difference is in our bot architecture post.
From leaderboard to follow set in five minutes
- Open /leaderboard with default sort.
- Apply the "Balanced" recipe above.
- Skim the top 30, opening 5–8 profile pages.
- For each, check: trades ≥ 100, max DD ≤ 15%, HHI ≤ 0.4, last active ≤ 72h, category histogram is sane.
- Pick 3–5 wallets that, between them, span at least 2 categories.
- Add to follow set. Set USDC range and category whitelist. Start engine.
The full setup walkthrough is in our auto-copy setup guide.
Frequently asked questions
How many wallets are on the Polymarket leaderboard?
Poly Syncer indexes 12,438 wallets, of which roughly 1,891–2,492 meet the 30-trade minimum and appear on the ranked leaderboard at any given refresh. The cohort fluctuates as wallets gain or lose qualifying samples in the rolling 30-day window.
How often does the leaderboard refresh?
Every 15 minutes. Each refresh recomputes the composite from a fresh 30-day rolling window. The exact last-refresh timestamp is shown at the top of the leaderboard.
Why does the same wallet sometimes drop 20 ranks in one refresh?
Usually because a major market the wallet held resolved against them, and the trade observation hit the return series. Composite is z-scored against the cohort, so a meaningful shift in one wallet's series can move it noticeably. Drops of 20+ ranks during quiet periods are rare; during major resolution waves they are normal.
Should I copy the #1 wallet?
Generally no. The top of the leaderboard is often a small-sample-size outlier with concentrated positions. A more durable approach is to filter to "trades ≥ 200, max DD ≤ 12%, HHI ≤ 0.3" and pick from the resulting list. The exact-rank-1 wallet is news; the top-50 set with proper filters is signal.
Does Poly Syncer manually curate the leaderboard?
No. The score is fully algorithmic. The only manual interventions are documented exclusion of wallets identified as the same operator's secondary addresses (de-duplication) and removal of clear self-trading clusters. Both are documented on /methodology.
Can I see the leaderboard via API?
Yes. The full ranked list and per-wallet inputs are available at /developers. Rate limits are documented; the schema matches the on-screen columns one-to-one.
What does an HHI score actually mean?
HHI is the sum of squared market shares of a wallet's USDC volume. A wallet that put 100% of its volume in one market has HHI = 1.0; a wallet evenly spread across 10 markets has HHI = 0.10. We treat HHI > 0.4 as flagged-concentrated. The glossary has the full definition.
Where to go next
If you have read this far, the next two articles in the series are building a smart money tracker — how we go from raw on-chain data to a ranked wallet set in the first place — and the auto-copy setup walkthrough that converts leaderboard browsing into actual mirrored fills. The full whitepaper covers the engineering stack end to end. When you are ready to start mirroring, the upgrade flow is at /dashboard/billing.