Reading a Polymarket wallet like a quant: Sharpe, drawdown, Kelly, and edge

Four numbers separate a copyable Polymarket wallet from a lucky one: its realistic post-fee edge, its risk-adjusted return, its maximum drawdown, and the Kelly fraction that ties them together. Raw profit-and-loss — the only number a leaderboard shows you — reveals none of them. A quant never reads a track record by how much it made; they read it by how the money was made, how much risk was carried to make it, and whether any of it is likely to repeat. Here is how to do the same to a wallet before you copy it.

Edge: what's actually repeatable?

Edge is the average amount a wallet makes per dollar staked, net of costs — the part of the record that comes from being right, not from being big or lucky. The number that matters to a copier is the realistic post-fee edge: gross edge minus the taker fees, slippage, and copy latency you pay and the leader didn't. A wallet's advertised headline return assumes none of those, so it overstates what a copier keeps by roughly 20–40%.

The base rate is harsh. Across the 1,649 actively-traded wallets under CopyGrade coverage, 74% have a negative post-fee edge and the median sits at −2.7% — losing money before a copier adds a cent of their own costs. Edge is also the one stat that survives manipulation least well, which is why it anchors the edge-authenticity sub-score: a high win rate can be manufactured or meaningless, but a durable positive edge across hundreds of resolved markets is hard to fake.

Risk-adjusted return: skill, or survived variance?

A risk-adjusted return (a Sharpe-style ratio) measures return per unit of volatility — how much the wallet made relative to how wildly its equity swung to make it. It exists because two wallets with the identical raw profit can have produced it in opposite ways: one from a steady, repeatable edge applied at sane size, the other from a single oversized bet that happened to land.

On a raw PnL chart those two wallets look the same. On a risk-adjusted basis they look nothing alike — and only one of them is likely to survive the next regime. This is the risk-adjusted-return sub-score's whole job: to dock the wallet that "won" by betting the farm and reward the one whose return came from being right repeatedly. A streak and a strategy produce the same headline; only the ratio tells them apart.

Maximum drawdown and recovery: the stretch you must survive

Maximum drawdown is the largest peak-to-trough fall in a wallet's equity — the worst losing stretch you would have had to sit through while copying it. It matters more than most copiers expect, because losses and the gains needed to undo them are not symmetric: the deeper the hole, the disproportionately larger the climb out.

The math is fixed: recovering a loss of L requires a gain of L ÷ (1−L). A 10% loss needs an 11% gain to get back to even; a 25% loss needs 33%; a 50% loss needs a 100% gain — you must double what's left; and a 75% loss needs 300%. This is why the drawdown-resilience sub-score reads the depth and recovery of a wallet's worst stretch, and why you size every wallet from its drawdown, not its best month. A wallet you can't hold through its worst week is one you'll abandon at the bottom — locking in the loss right before the recovery.

Kelly: how your edge sets your size

The Kelly criterion is the bet size that maximises the long-run growth rate of a bankroll, given an edge and the odds. In plain terms, it scales your stake to how big your edge is and how favourable the price is — bet more when the edge is large and the odds are long, less when the edge is thin. It is the bridge between "this wallet has an edge" and "so how much do I put behind it."

Two things make Kelly a due-diligence tool, not just a sizing formula. First, full Kelly is the theoretical maximum-growth bet and is brutally volatile — it routinely tolerates 50%+ drawdowns, which collide directly with the recovery arithmetic above, so most disciplined copiers stake a half or quarter of it. Second, over-betting even a genuine edge still busts you: past the Kelly fraction, additional size lowers long-run growth and raises ruin risk. So a wallet's edge doesn't just tell you whether to copy — it caps how much you can copy before risk overwhelms the return. The position-sizing guide turns this into concrete rules.

Sample size and independence: the metric under the metrics

Sample size is how many resolved, independent markets a wallet's record rests on — and it governs whether any of the four numbers above mean anything at all. An edge measured over fifteen correlated election bets is a coin flip dressed as a strategy; the same edge over several hundred uncorrelated, resolved markets is a signal.

Independence matters as much as count. Ten wallets that all bet the same Iran market are one bet held ten times, not ten data points — and a wallet whose entire record is a handful of correlated event-week spikes has a sample of roughly one, no matter how many trades it logged. Small or correlated samples are why a flashy short record should be read with suspicion: with too little independent data, edge, Sharpe, and drawdown are all just noise wearing a suit.

How these map to a Copy Score

Each metric answers one question, and each feeds one part of the Copy Score. Read them together — a wallet that clears one and fails another is not copyable on the strength of the one it passed.

Hold a wallet to all of them at once and the field thins fast: only about 1.3% of active wallets clear every test. That is the bar the per-wallet Copy Verdict checks for you, and the Copy Simulator lets you stress-test a candidate's drawdown and post-fee edge under your own stake and latency before any money moves. The same lens explains why most winning records don't survive copying: they were never measured on these numbers in the first place.

CopyGrade is analysis-only — it never executes trades or holds funds, and a Copy Score is a documented research opinion, not financial advice.