We re-scored 1,649 Polymarket wallets. It got worse.
This is the third edition of our recurring report on the state of Polymarket copy trading, and the honest summary is that it got worse. We re-scored the full universe under CopyGrade coverage — 1,649 actively-traded wallets, up from 1,297 in June and 385 in the first edition — and 60% now carry a farming flag, up from 53%. The median fee-adjusted edge deteriorated from −0.8% to −2.7%. Just 22 wallets — 1.3% — pass every test a copier should apply. And unlike last edition, the deterioration is not explained by broader sampling: the wallets we were already covering in June got dirtier in place.
A note on the sample: 1,649 wallets that trade actively enough to be scored, snapshotted in early July 2026, each graded from its public Polymarket trade history by the same Copy Score model as every prior edition. The model's weights and thresholds are unchanged since June — we published a 30-day validation of the score against realized outcomes on July 1 and deliberately changed nothing. Where the instrument's reach grew anyway, we say so below.
The headline numbers, June vs. July
| June 2026 (1,297) | July 2026 (1,649) | |
|---|---|---|
| Median Copy Score | 35 | 30 |
| Farming-flagged | 53% | 60% |
| — high-confidence (severe) | 43% | 51% |
| Negative fee-adjusted edge | 70% | 74% |
| Median fee-adjusted edge | −0.8% | −2.7% |
| Worth copying (all tests at once) | 28 (2.2%) | 22 (1.3%) |
Every line moved against the copier. The share of wallets with any genuine post-fee edge shrank, the farming-flag rate rose seven points, and the copyable tail thinned from 28 wallets to 22 — even though the scored population grew by 352. Seventy-three percent of the scored set now sits below a Copy Score of 50.
One number makes the copier's problem concrete: 316 wallets currently show a positive headline edge — and 238 of them, three in four, carry a farming flag. The records that look most worth copying are, at this snapshot, mostly records our model assesses as manufactured-risk. A good-looking history is not evidence of a good target; at these base rates it is closer to the opposite.
The vetting funnel: 1,649 in, 22 out
Each row below is how many wallets pass that single test; the last row is how many pass all four at once.
| Test | Wallets passing | Share |
|---|---|---|
| Scored | 1,649 | 100% |
| Enough recent history to judge (≥20 trades / 90d) | 1,300 | 79% |
| Farming-clean | 659 | 40% |
| Fee-adjusted edge above 1% | 187 | 11% |
| Copy Score ≥ 75 | 23 | 1.4% |
| All four simultaneously — worth copying | 22 | 1.3% |
In June, 48 wallets (3.7%) were simultaneously clean, active, and genuinely profitable even before applying the score bar. In July that intersection is down to 35 wallets (2.1%), and the full bar passes 22. The funnel isn't narrowing because we raised it — the tests are identical — it's narrowing because fewer wallets clear the same gates.
Is the rise real, or is the model just seeing more?
Last edition we attributed the flag-rate rise honestly to composition: tripled coverage had pulled in the leaderboard cohort, which is dirtier than the field. This time composition points the other way, so we split the population and measured.
| Cohort | Wallets | Farming-flagged | Median fee-adj. edge |
|---|---|---|---|
| Already covered in June | 1,308 | 62% (was 53% across a near-identical set) | −2.4% |
| Newly covered since June | 341 | 54% | −8.0% |
The newly-covered wallets are cleaner on flags than the incumbents — so the seven-point rise happened in place, inside the set we were already watching. Three things plausibly drove it, and we can only partially separate them:
- The instrument's reach grew. Since June we scaled cross-wallet cluster detection to evaluate far more candidate wallet pairs, and improved on-chain funder resolution — both changed in public, with dated notes. Linkage signals the June run structurally could not see are now visible. That raises measured flag rates without any wallet changing its behaviour.
- Histories deepened. Farming signatures are behavioural patterns in the trade history; a longer tape per wallet gives the detectors more to find. Coverage depth grew for nearly every incumbent wallet between editions.
- Behaviour may genuinely be shifting. The still-unshipped POLY airdrop continues to reward manufactured volume, and mid-2026 is the highest-volume era in Polymarket's history. We flag this as plausible, not proven.
We can't fully apportion the rise between those three, and we won't pretend to. What we can say: the ruler didn't get stricter — no threshold or weight moved — but its resolution improved, and the picture at higher resolution is worse.
The edge deterioration is harder still to attribute. Nothing changed in how fee-adjusted edge is computed; the same wallets simply grade out worse on longer histories (−0.8% → −2.4% median for the incumbent cohort), and the newly-covered cohort arrives at a brutal −8.0%. The uncomfortable reading, consistent with what a $67B academic study found about who wins on Polymarket, is that short windows flatter almost everyone, and time un-flatters them.
The leaderboard cut: one copyable wallet
The wallets currently on Polymarket's public profit and volume boards — 159 under coverage at this snapshot; the roster churns as windows roll — remain worse than the field, extending last edition's finding:
| Full set (1,649) | Current leaderboard (159) | |
|---|---|---|
| Median Copy Score | 30 | 30 |
| Farming-flagged | 60% | 69% |
| — severe | 51% | 59% |
| Median fee-adjusted edge | −2.7% | −9.7% |
| Worth copying | 22 (1.3%) | 1 (0.6%) |
One wallet. The public boards rank by raw profit, raw profit rewards exactly the behaviours our model penalises, and at this snapshot precisely one board name survives every check. If your shortlist is the leaderboard, your base rate is one in 159 — see the graded leaderboard for the rank-versus-reality gap on today's board.
What this means if you copy
The base rates are the strategy. At 1.3%, picking a wallet by eye is a 98.7%-failure lottery, and the failure modes are asymmetric — a farmed record doesn't underperform, it turns its copiers into exit liquidity.
- Vet before you automate. The pre-automation checklist exists because most candidates fail it.
- Read edge net of costs. The headline return is a ceiling you'll never reach; the median wallet is negative before you add slippage.
- Treat a clean-looking record as unproven, not clean. Three in four positive-edge records carry a flag this month. Check the behavioural signatures, not the PnL.
- Re-check what you already copy. These numbers moved in four weeks. Alerts exist so a deteriorating target reaches you before your bot keeps mirroring it.
How we got these numbers
Point-in-time snapshot, early July 2026: 1,649 actively-traded wallets under CopyGrade coverage, re-scored within a single sync cycle by the current Copy Score model from public Polymarket trade history. Edge figures are net of modelled trading costs; we quote medians throughout, which the long negative tail doesn't distort. Farming flags are algorithmic risk assessments — our documented opinion from disclosed public data, not verified conduct or a statement of fact about any trader, and each wallet's page carries a dispute path. The model changes in public, with dated notes; this edition reports the same weights validated — and found wanting, and kept anyway pending more data — in our published null result. Prior editions: June 8, June 2. Not financial advice.
CopyGrade is analysis-only — it never executes trades, holds funds, or custodies keys, and a Copy Score is a documented research opinion, not financial advice.