Across 12,438 Polymarket wallets we scored over a 90-day window, the top 10 returned a median 87% with a median 30-day Sharpe of 3.1, and they share four traits: tight position sizing, category specialization, off-peak timing, and a hit rate that beats implied odds by 6–11 percentage points. This post unpacks the data, names patterns, and shows the truncated addresses of the top Polymarket traders you can follow on the Poly Syncer smart money tracker leaderboard.
How we built this dataset
We scored 12,438 wallets that placed at least 30 trades in the 90-day window ending May 1, 2026. Scores follow the composite described in our wallet-scoring methodology — a Sharpe-weighted blend of risk-adjusted return, edge-adjusted hit rate, and drawdown. Outlier-driven wallets were winsorized via a Hampel/MAD filter (see MAD on Wikipedia). The full schema lives on /methodology.
Top 10 wallets by composite score
Addresses are truncated for readability. ROI and Sharpe are 30-day. "Primary cat" is the category accounting for the largest share of the wallet's USDC volume.
| # | Wallet | ROI 30d | Sharpe | Trades | Edge HR | Primary cat |
|---|---|---|---|---|---|---|
| 1 | 0xA31F…9C24 | +118% | 3.9 | 214 | +0.11 | Politics |
| 2 | 0x7E0B…4F12 | +102% | 3.6 | 388 | +0.09 | Sports |
| 3 | 0x5142…88D7 | +96% | 3.5 | 147 | +0.10 | Crypto |
| 4 | 0xB9C1…01AE | +89% | 3.3 | 312 | +0.08 | Earnings |
| 5 | 0x33F4…7B62 | +87% | 3.2 | 96 | +0.07 | Geopolitics |
| 6 | 0xCE82…2A19 | +86% | 3.0 | 421 | +0.09 | NBA |
| 7 | 0x12AA…F4D8 | +82% | 3.1 | 178 | +0.10 | Politics |
| 8 | 0x9D77…C310 | +78% | 2.9 | 263 | +0.06 | Tech |
| 9 | 0x6B0E…A551 | +74% | 2.8 | 109 | +0.08 | Fed Rates |
| 10 | 0x44D2…88FE | +72% | 2.7 | 501 | +0.05 | Soccer |
Two observations jump out before any analysis: edge-adjusted hit rates cluster between +0.05 and +0.11 (i.e., these wallets beat implied probability by 5–11 points consistently), and trade counts vary by 5× — some leaders are concentrated specialists, others run high-frequency books.
Pattern 1: category specialization beats generalism
Of the top 10, eight derive 60%+ of their volume from a single category. The two generalists (#4 and #8) have the lowest edge-adjusted hit rates in the cohort. Across the broader top-100, the median wallet has 71% of volume in its top category. The top-1000 generalists average 38%. Concentration correlates with edge.
This is intuitive. A trader who reads earnings transcripts and trades earnings-night markets is using domain knowledge that does not transfer to NBA player props. The data agrees: when we restrict each top-10 wallet's track record to only their primary category, edge-adjusted hit rates rise by an average of 1.8 percentage points.
Pattern 2: politics has the highest Sharpe; sports the highest volume
Aggregating the whole 12,438-wallet cohort:
| Category | Median Sharpe (top-100) | Median trades/wallet | Volume share |
|---|---|---|---|
| Politics | 2.8 | 142 | 31% |
| Sports (incl. NBA, Soccer) | 2.1 | 288 | 36% |
| Crypto | 2.4 | 119 | 11% |
| Earnings | 2.6 | 87 | 8% |
| Geopolitics | 2.3 | 61 | 6% |
Politics has the highest median top-100 Sharpe (2.8) because resolution timelines are long and informed traders have time to accumulate position. Sports has the most volume because the resolution cadence is daily and traders churn through more contracts. Both are real edges, but they live in different parts of the strategy space. Our categories explained post covers all 25.
Pattern 3: time-of-day (UTC)
Plotting top-100 fill timestamps in UTC:
- 03:00–06:00 UTC: 18% of all volume but 34% of profitable closed positions in the cohort. This is late-night Asia / early Europe; books are thinner and informed orders move price.
- 14:00–19:00 UTC: 41% of volume but only 28% of profitable closes. US-business-hours flow is louder and noisier.
- 22:00–01:00 UTC: resolution-heavy window for North-American sports; high turnover, average edge.
The rough takeaway: the best 30-day Sharpe wallets are disproportionately active when the rest of the market is asleep.
Pattern 4: position-size discipline
Average position size as a percentage of bankroll for the top 10 is 3.4%. The median across the bottom-1000 (negative ROI) is 11.7%. Re-said: losing wallets bet 3.4× bigger as a share of bankroll than winning wallets. Kelly would prescribe even less; we cover the math in detail in our Kelly post.
The most consistent finding in our 90-day dataset is not what the top wallets bet on. It is how much they bet.
Pattern 5: holding period and exit discipline
Median holding time across the top 10 is 14 hours; median across the bottom-1000 is 4.6 days. Top wallets close losing positions decisively; underperformers ride them to resolution and accept the binary outcome. Concretely, in our 90-day sample:
- Top-100 wallets closed 71% of trades before resolution; bottom-1000 closed 31%.
- Top-100 average loss-on-loser was 18% of position; bottom-1000 averaged 64% (i.e., losers held to expiry tend to go to zero).
- The ratio of average win to average loss for the top 10 is 2.3; for the bottom-1000 it is 0.6.
This shows up cleanly in copy-trading data: when you mirror a top wallet, you are inheriting both their entries and their exits. The exits are arguably more valuable. A leader who buys well but sells badly is a very different proposition from one who does both, and the trade-history tab on each profile lets you inspect exit behavior before following.
Pattern 6: market-liquidity preference
The top-10 wallets executed 78% of their volume in markets with >$50,000 of depth at top-of-book. Below that depth, slippage on a $250 mirror is structurally meaningful (we measure 1.4–2.1 cents median tracking error in <$5,000-depth markets vs 0.3–0.6 cents in deep books). For copy traders this means: the leaders worth following are usually the ones already trading in books where you can mirror at scale without paying excess slippage. The Poly Syncer executor enforces a default liquidity floor (configurable) precisely so mirrors do not fire into thin books.
What you can take away if you trade manually
- Specialize. Pick one category. Build a model. Ignore the others until that one prints.
- Cap positions at 3–5% of bankroll. The data is unambiguous.
- Trade off-peak. If you can stomach 04:00 UTC entries, you are competing against thinner books.
- Track edge-adjusted hit rate, not raw win-rate. Without this metric you cannot tell skill from $0.95-favorite cherry-picking.
How the top-10 distribution compares to the broader cohort
It helps to anchor those top-10 numbers against the rest of the universe. Cohort summary statistics across all 12,438 scored wallets in the 90-day window:
| Percentile | 30d ROI | 30d Sharpe | Edge HR | Max DD |
|---|---|---|---|---|
| 99th | +72% | 2.7 | +0.06 | −9% |
| 95th | +34% | 1.9 | +0.03 | −14% |
| 75th | +8% | 0.9 | 0.00 | −22% |
| 50th (median) | −3% | 0.1 | −0.02 | −31% |
| 25th | −14% | −0.6 | −0.04 | −42% |
The median wallet loses 3% over a 30-day window and runs a 31% drawdown to do it. That is the universe you are competing against if you trade manually, and it is the noise you are filtering through when you screen the leaderboard. The Pareto distribution of edge in this market is extreme: the top 2% account for an outsized share of total profit, and the bottom 50% are net negative-sum after slippage. Picking from the right tail with sample-size discipline (≥100 trades) and concentration discipline (HHI < 0.4) is the entire game.
What separates the top 2% from everyone else
Re-reading the patterns in aggregate, the top 2% of wallets (by composite score) share four observable, reproducible behaviors:
- Specialization — one category accounts for >60% of volume. Median across the top 100 is 71%; median across the bottom 1,000 is 38%. The right tail is overwhelmingly specialist.
- Sizing discipline — average position is 3.4% of estimated bankroll. Bottom 1,000 averages 11.7%. The simplest, most-replicable lesson in the dataset.
- Off-peak activity — 34% of profitable closes occur in the 03:00–06:00 UTC window, despite that window holding only 18% of total volume. Thinner books, less informed flow on the other side.
- Edge-adjusted hit rate — +0.05 to +0.11. Beating implied probability by 5–11 percentage points consistently is the single best forward-looking predictor of next-window Sharpe in our data, more so than past Sharpe itself.
If you mirror three wallets that each hit those four bars, with category diversification across them, you have effectively assembled a small synthetic fund whose risk profile is dominated by professional-quality decision-making. That is the actual product of copy trading done well, and it is what the leaderboard's filtering and the executor's gates are designed to make routine. The data does not promise returns — nothing in this universe does — but it does say that the configuration above is, repeatedly, where reproducible edge has lived in the prior 90 days.
The faster path: follow them
If you do not want to build a model from scratch, the entire top-10 list above is followable on the Poly Syncer leaderboard in a single click each, with your own size caps and stop-loss applied automatically. The free view-only tier lets you study every wallet on that list before paying; Pro at $299/month adds 250 mirrored wallets and premium-RPC execution.
A note on what this dataset can and cannot tell you
Three caveats to internalize before extrapolating. First, 90 days is a single window; election cycles, earnings seasons, and crypto regimes all shift the relative attractiveness of categories materially. Second, all numbers above are gross of any execution slippage when copying — Poly Syncer mirror tracking error of 0.3–0.6 cents per fill in deep books typically subtracts 4–9% from a leader's headline 30-day return when copied at scale. Third, the cohort is a survivor population — wallets that blew up and stopped trading are not in the universe, and that bias understates the variance of returns in this space. Treat the patterns as descriptive of a real distribution, not predictive guarantees.
Frequently asked questions
Do these rankings change frequently?
Yes. The leaderboard refreshes every 15 minutes; rank changes within the top 10 are typical between weekly snapshots. The cohort of "elite" wallets (top 100) is more stable — Spearman ρ between consecutive 15-min refreshes averages 0.92.
Is this data survivorship-biased?
Partially — we only score wallets active in the 90-day window. Wallets that blew up and stopped trading are not present. We discuss this honestly in the risk-management post.
Can I follow several of these at once?
Yes. Pro allows up to 250 mirrored wallets, Elite is unlimited. We recommend 2–3 with low return correlation rather than ten at once — correlated leaders amplify drawdowns.
Where do the truncated addresses come from?
They are real on-chain Polymarket wallets selected by the composite score on the snapshot date; we publish only the truncation here for readability. The full addresses appear on each wallet's profile page on the leaderboard.