We Tracked 6,792 Polymarket Wallets for 8 Months. Only 1% Stayed Good.

Every copy-trading guide tells you to find a top trader on the Polymarket leaderboard and follow them. Almost none of them ask the obvious follow-up question: does the trader who is on the leaderboard today still belong there next month? We have been snapshotting the Polymarket leaderboard daily since December 2025. Across 6,792 distinct wallets and roughly 240 daily snapshots, the answer is uncomfortable — the leaderboard is mostly churn, and the single metric everyone screens on tells you the least.

What we tracked, and what this data is not

Since 26 December 2025 we have captured the Polymarket leaderboard once a day and recorded every wallet that appears, along with its P&L, volume, win rate, position history and how many separate days it has shown up. As of 23 August 2026 that is 6,792 distinct wallets. One wallet has appeared 220 times. Most have appeared a handful.

One caveat matters more than any finding below, so it goes first. This pool is sourced from the leaderboard, which means it is already the top of the market. These are not "Polymarket traders" in general — they are traders who were good enough, or lucky enough, to surface on a public ranking at least once. Any statement here about profitability is a statement about the winners' bracket, not the field. That is a real limitation, but it is also precisely the population a copy trader is choosing from, which is what makes it worth measuring.

Finding 1: the leaderboard is a turnstile, not a ranking

We measure consistency as the share of observed days a wallet actually held a leaderboard position. If the board were a stable ranking of genuinely superior traders, consistency would cluster high. It does the opposite.

Seven wallets out of 6,792 held their place on four days in five. Three quarters of the pool showed up, posted a number good enough to rank, and then largely vanished from the board.

This is what you would expect if a large share of leaderboard appearances were driven by variance rather than durable skill — a good week on a concentrated position is enough to rank, and reversion does the rest. It does not prove that the traders are unskilled. It does mean that "they're on the leaderboard" carries far less information than the phrase implies, and that a screenshot of today's top ten is close to worthless as a copy list.

Finding 2: win rate is the most quoted and least useful number

Win rate is the metric every copy-trading post leads with, and it is the easiest one to read backwards. Here is median performance by win-rate bucket across the pool.

Read the P&L column first. A wallet winning fewer than half its positions had a median profit of $44,653. A wallet winning more than 80% of them had a median profit of $48,136. Between those two extremes, median profit does not move in any meaningful direction — the 60–70% bucket actually posts the highest figure in the table.

Now read the last two columns together, because that is where the real story is. ROI climbs steadily with win rate (1.7% to 5.1%) while median volume falls by half (from $2.32M to $1.15M). High win rate is not a marker of a better trader. It is a marker of a different style — smaller, more selective, favourite-heavy positions that resolve in your favour most of the time and pay little when they do. Low win rate with equal profit means the opposite: bigger books, more losers, and larger payoffs on the winners.

Neither style is wrong. But if you screen for "80%+ win rate" believing you are filtering for skill, you are actually filtering for position size and price selection, and you will systematically exclude a group of traders making the same money a different way. If you then copy an 80% wallet at your own position sizes, you inherit their hit rate without their sizing discipline — which is a different bet from the one you thought you were placing.

Finding 3: stack four basic screens and 99% of the pool disappears

Individually, none of the following is a demanding requirement. Together they are brutal.

• At least 90 days of track record. 52.3% of the pool clears this. Another 12.5% have under 14 days of history — not enough to distinguish an edge from a streak.
• Consistency of 50% or better. 1.9% clear this, per the table above.
• Actually profitable on realised P&L.
• Positive risk-adjusted return — the trader is being paid for the volatility they take on.

Wallets that satisfy all four at once: 69 out of 6,792, or 1.0%.

That number is the practical takeaway of the whole study. The scarce thing in copy trading is not finding a profitable wallet — in a pool skimmed off the leaderboard, most are profitable by construction. The scarce thing is finding a wallet that is still there in three months, having stayed profitable in a way that was not mostly variance. Roughly one in a hundred leaderboard wallets currently looks like that.

What this changes about how you pick someone to copy

Stop treating the leaderboard as a shortlist

It is a discovery surface, not a ranking of who is worth following. Use it to generate candidates, then screen them on history rather than on the snapshot that put them there. If the only evidence for a trader is that they are in today's top ten, you have almost no evidence.

Weight track record above headline returns

A 90-day history is the cheapest filter in the list and it removes nearly half the pool on its own. The 12.5% of wallets with under two weeks of data should not be copyable at all — there is no way to tell a system from a streak at that sample size.

Read win rate as a style tag, not a quality score

Ask what the number implies about position size and price selection. A 55% win rate on large volume and a 85% win rate on small volume can be the same trader quality expressed two ways. What you actually need to know is whether your bankroll and temperament match the style you are about to inherit. Our guide to position sizing when copying a whale covers how to translate someone else's trades into sizes that fit your account.

Check whether they are still trading

A wallet with a superb historical record and no recent activity is a museum piece. Recency is not in the four screens above, and it should be in yours.

Method and limitations

Data is drawn from daily Polymarket leaderboard snapshots between 26 December 2025 and 23 August 2026, covering 6,792 distinct wallets. Consistency is the share of observed days on which a wallet held a leaderboard position. P&L, ROI, volume and win rate come from on-chain position and trade history at the time of each analysis.

Three honest limits. Selection bias: as stated above, this is the leaderboard population, not the Polymarket user base — profitability rates here say nothing about the average trader. Survivorship within the window: wallets discovered late in the period have fewer observed days, which compresses their possible consistency. Backward-looking: every figure describes what these wallets have already done. None of it is a forecast, and we would encourage scepticism toward anyone — us included — who presents a historical score as a prediction of future returns. The value of a screen is that it removes wallets you can demonstrate you should not copy, not that it identifies wallets guaranteed to keep winning.

The short version

Of 6,792 wallets good enough to reach the Polymarket leaderboard at least once, seven held their spot four days in five, 128 held it half the time, and 69 passed a plain four-part screen. Win rate, the metric that dominates every copy-trading discussion, did not separate the profitable from the unprofitable — it separated the cautious from the aggressive. If you take one operational change from this: screen on track record and consistency first, and treat today's leaderboard as a list of candidates rather than a list of answers.