We scored 385 Polymarket wallets. Only 13 were worth copying.
We scored 385 of Polymarket's most actively-traded wallets — the kind that surface near the top of a volume or profit leaderboard, the ones a copy-bot would actually point at. Thirteen of them — 3.4% — cleared the bar for a wallet genuinely worth copying: enough recent history to judge, no farming signatures, a realistic post-fee edge, and a strong Copy Score. The other 96.6% failed at least one test. This is what the full distribution looks like, and why the leaderboard hides it.
A note on the sample: these are 385 high-volume wallets under CopyGrade coverage, surfaced from Polymarket's busiest markets — not a random or complete census of the platform's hundreds of thousands of accounts. It's a deliberate look at the actively-traded set most likely to be copied. All figures are a snapshot from early June 2026 and will move as coverage grows and the model is recalibrated in public.
The leaderboard ranks the wrong number
Polymarket's leaderboards rank by raw profit. Raw profit is the one statistic that tells you almost nothing about whether copying a wallet makes you money: it rewards account size, reckless sizing, and dumb luck, and it's computed on the leader's own fills — their prices, their timing, zero fees. A copier arrives one step behind, pays taker fees on the way in and out, and eats slippage and latency. So "up the most" and "worth copying" are barely related — and once you score the wallets on what a copier would actually keep, most of the leaderboard evaporates.
Across the 385 wallets, the median Copy Score is 46 out of 100. Only 14 wallets (3.6%) score 75 or above — the "strong copy candidate" band — while 219 (57%) land below 50, in the Avoid band. The distribution is not a bell curve centred on "pretty good." It's a long tail of wallets that look fine on a profit ranking and fall apart the moment you ask the harder question.
Most "top" wallets have no edge to copy
Strip the headline number back to a fee-adjusted edge and the picture is stark. The median wallet sits slightly negative — a −0.6% 90-day edge net of its own trading fees, before a copier adds a cent of slippage or latency. Only 88 wallets (23%) have a positive fee-adjusted edge at all. Tighten that to a realistic post-fee, post-latency edge above 1% — a low bar for "worth the trouble of copying" — and just 55 wallets (14%) clear it.
The sub-scores say why. The median wallet's risk-adjusted-return factor is 19 out of 100 and its edge-authenticity factor is 46 — meaning whatever edge exists tends to read as variance rather than repeatable skill, carried by sizing rather than by being right. A high raw-profit number on a wallet like that is a streak, not a strategy. The latency tax — fees, slippage, and the delay between the leader's fill and yours — then shaves whatever is left; but the more striking finding in this sample is that for most wallets there's simply very little edge there to tax.
Farming is the norm, not the exception
The reason due diligence isn't optional: a large share of high-volume wallets aren't just unprofitable to copy — they're built to profit from copiers.
| Farming level | Wallets | Share |
|---|---|---|
| Clean | 243 | 63% |
| Watch — emerging signals | 32 | 8% |
| Severe — high-confidence | 110 | 29% |
37% of the wallets carry a farming flag, and 29% carry a high-confidence (severe) one — iceberg accumulation, self-trade wash, decoy clusters, stealth merges, pump-and-dump, copy-bait. These are the signatures of a wallet that builds a clean-looking record quietly, attracts copiers, and then exits into them.
The overlap with "looks profitable" is the dangerous part. Of the 88 wallets with a positive headline edge, 31 — more than a third — are farming-flagged. 21 wallets carry a severe flag and a positive headline edge: precisely the profile that ranks well and reads as a great copy target, and is set up to turn its followers into exit liquidity. The leaderboard can't see any of this, because farming is a behavioural pattern in the trade history, not a number in the PnL column.
The copyability funnel: 385 → 13
Here's how the 385 wallets fare against each test a copier should actually apply. Each row is how many of the full set pass that single criterion; the final row is the count that passes all four at once.
| Test | Wallets passing | Share |
|---|---|---|
| Scored | 385 | 100% |
| Enough recent history to judge (≥20 trades / 90d) | 277 | 72% |
| Farming-clean | 243 | 63% |
| Realistic post-fee edge above 1% | 55 | 14% |
| Copy Score ≥ 75 | 14 | 3.6% |
| All four simultaneously — copyable | 13 | 3.4% |
Loosen the bar — drop the Copy Score requirement and just ask for clean, genuinely profitable, and active — and you still only reach 23 wallets (6%). Either way, the actively-traded Polymarket population is overwhelmingly not copyable, and the handful that are don't stand out on a profit leaderboard. Finding them is the entire job.
What this means if you copy
The leaderboard is a list of candidates, not a list of answers — and the base rate is brutal enough that picking from it at random is a losing game. Before you point a bot at any wallet, three checks separate the 3.4% from the rest:
- Post-fee edge, not raw profit. Ask what a copier would actually keep after fees, slippage, and latency — the headline number is a ceiling you'll never reach.
- Farming forensics. A clean-looking record is exactly what a farmer builds. Check the trade history for the behavioural signatures, not just the PnL.
- Risk-adjusted return. A streak and a strategy produce the same headline; only one survives the next regime.
That's the work the Copy Score does on every wallet — and the per-wallet Copy Verdict shows the post-fee edge, the farming check, and the risk-adjusted breakdown in one place, so you can see which side of the 3.4% a wallet sits on before you commit capital.
How we got these numbers
The figures above are a point-in-time snapshot (early June 2026) of 385 actively-traded Polymarket wallets under CopyGrade coverage, drawn from the platform's highest-volume markets — a focused sample of the set most likely to be copied, not a complete census. Every wallet is scored from its public Polymarket trade history by the current Copy Score model, recomputed on each sync; the model itself is tuned against real outcomes and changes in public, with a dated note, so these numbers will shift as it's recalibrated and as coverage widens. CopyGrade is independent and analysis-only — it never executes trades or holds funds. Not financial advice.