How to vet a Polymarket trader: the full checklist
Vetting a Polymarket trader means answering five questions from their public trade history — is the edge real, what risk produced it, how deep do the losses run, where does the edge live, and is the record honest — and refusing to copy until all five clear. Polymarket's own copy-trading guidance stops at qualitative tips; this is the quantitative version, with the thresholds we use and the base rates that show why each test earns its place. The short editorial version is our original checklist post; this guide is the full reference.
The base rates first, because they set expectations. In our July 2026 snapshot of 1,649 actively-traded wallets, here's the share passing each test individually:
| Test | Share of active wallets passing |
|---|---|
| Enough recent history to judge (≥20 trades / 90d) | 79% |
| No farming flag | 40% |
| Realistic post-fee edge above 1% | 11% |
| Worth copying (passes everything at once) | 1.3% |
Six in ten wallets you'll consider fail the honesty test alone, and only about one in 75 clears everything. Vetting isn't paranoia; it's the base rate.
1. Is there enough history to judge at all?
Before judging quality, judge quantity. Our floor is 20 trades in the last 90 days — below that, every other statistic is noise, which is why thin wallets aren't even scored against the full model. But the floor is a floor: for the edge tests below you want hundreds of resolved trades across genuinely independent markets. Twelve correlated bets on one election week tell you almost nothing; the same hit rate across three hundred markets and several categories is an actual claim.
Red flags at this stage: a record shorter than one market regime, profits concentrated in a single event, or a handful of oversized winners carrying the whole PnL.
2. Is the edge real after copying costs — not before?
The number to demand is the realistic post-fee edge: what a copier would keep per unit staked after fees, slippage, and copy latency — not the headline return, which a copier typically undershoots by 20–40%. Our bar is a realistic edge above 1%; only 11% of active wallets clear it, and the median wallet sits at −2.7% — negative before you add a single copier-side cost.
Two practical sub-checks:
- Speed-dependence. If the edge concentrates in fills that were profitable only moments after they happened, it belongs to whoever's first — and that is never the copier.
- Turnover. High turnover multiplies every per-trade cost. A modest edge on low turnover often copies better than a larger edge that churns.
3. What risk produced the return — and what's the worst drawdown?
Return per unit of risk is what repeats; raw return is what got lucky. Compare wallets on a Sharpe-style ratio (return over the volatility of returns) rather than the headline, and then look at the maximum drawdown — the worst peak-to-trough loss the wallet actually lived through, and how long the climb back took.
The test is personal: scale that drawdown to the capital you'd deploy. If the wallet's worst stretch would have cut your stake by a third, you need to have decided in advance that you'd sit through it — because statistically, you'll copy through one. A shallow-drawdown wallet with a lower headline beats a cliff with a better one.
4. Where does the edge live?
Edge is almost never uniform across categories. A wallet printing money on NBA totals may be lighting it on fire on geopolitics; naive copying mirrors both. Break the wallet's PnL down by category and check three things: whether profit concentrates somewhere specific, whether the current mix of its trading matches where its edge historically lived, and whether your bot can filter to just those categories.
A wallet that recently drifted from its profitable category into new territory is mid-experiment — you'd be funding the experiment.
5. Is the record honest? (the farming check)
The most dangerous wallet on Polymarket is not the bad trader — it's the good-looking one built to be copied. About six in ten active wallets (60%) carry a farming flag, and 51% a severe one; on the leaderboard itself it's 69%. The signatures are specific, on-chain, and invisible to a PnL column:
- Iceberg accumulation — positions built in small slices (sub-$200 is the classic size) to stay under copy-bot alert thresholds before the record is advertised.
- Self-trade wash — manufacturing volume and win-rate against one's own orders; an independent Columbia study estimates about a quarter of Polymarket's all-time volume is consistent with wash trading.
- Decoy clusters — losses parked on linked sub-wallets so the flagship looks clean; visible in funding-graph analysis, not in the wallet's own history.
- Invisible exits — merging YES+NO back to USDC to exit without a visible sale while copiers hold the bag.
This check vetoes the others: a wallet running these patterns is disqualified however good the rest of the file looks, because the better it looks, the more deliberate the construction. It's also the check that's effectively impossible to run by hand at shortlist scale — which is why our detectors automate it and a strong finding caps the Copy Score outright.
Running the checklist
Manually: pull the wallet's history from Polymarket's public data, compute post-fee returns FIFO, bucket by category, scan the funding graph — a day's work per wallet, honestly done. Or let the machine do it: the Copy Verdict runs all five checks on any wallet and shows the result as one score with the evidence underneath, and the Copy Simulator replays the copy under your capital, fees, and latency before a dollar moves. Either way, the discipline is the same — and the one-in-fifty base rate is why it pays.
CopyGrade is analysis-only — it never executes trades or holds funds. Scores are our documented opinion from public data, recalibrated in public. Not financial advice.