Who is this trader, really?

Paste any Polymarket wallet. Clause reads its trade history, works out what kind of trader it is, tests whether the record shows a real edge, and scores whether that edge would survive you copying it.

Reads the whole visible history, values every settled position, and walks the books they trade. Ten to thirty seconds. No key, no signup.

try

Why most copy-trading scores are wrong

The failure

Profit is the least useful of the three questions

Leaderboards rank wallets by realised profit. Three things sit between a profitable wallet and a profitable copier, and profit is not one of them.

Is the sample bets or fills? Twenty trades inside one football match are one draw, not twenty. Counting fills overstates significance by roughly the square root of the cluster size.

Is the trader even directional? A market maker quoting both sides has no direction to follow. Scoring it against a directional test reports a failure where there was never an attempt.

Does the price still exist? You trade after their print. If the market moves away in those seconds, you buy the same contract worse than they did, every time.

Each is measured here, and each can veto the verdict.

The measurement

Edge per share, not dollars

Every fill is already a prediction: buying at 0.34 asserts the outcome is worth more than 34 cents. When the market settles, that fill scores outcome − entry — the profit a copier makes on one share.

Dollar profit carries the variance of position size on top of the variance of being right. One large conviction bet swamps fifty small ones, so a bootstrap on profit is dominated by how much they staked rather than how well they chose — which is why that test almost never clears zero, on anyone.

Removing the size term makes a real edge visible in tens of markets instead of hundreds. It also refuses to reward staking, which belongs in the capacity leg and not smuggled into the edge test.

Averaged over markets, bootstrapped over markets.

What the score is made of

Every component publishes its own maximum and the reason it fell short. A score you cannot take apart is a number to be believed rather than checked, and this is a verification product.

Edge — 40

Scored on the worst case of the 90% interval, not the mean, because the lower bound is what a copier can rely on.

Sample — 20

Resolved markets, on a square-root scale. Evidence accumulates with the square root of the sample, so the scoring does too.

Independence — 15

One market supplying most of the edge is one correct call, which a copier joining today cannot repeat.

Followability — 15

Drift after their print, charged when it moves against a late copier and credited when it moves for them.

Capacity — 10

Whether your size actually fills in the markets they trade, at the depth on the book right now.

The caps

A non-directional strategy is capped at 25 and labelled, not failed. A record under 30 resolved markets cannot reach the top band however clean it looks.