← All posts · July 1, 2026 · 6 min read

Why Polymarket win rates are misleading

Wallet trackers put one number front and center: 87% win rate. Sounds like a machine that prints money.

It's the first number everyone looks at, and it's close to the last number that should decide anything.

What win rate actually measures

On Polymarket, win rate means one thing: how often the side the trader picked ended up winning. That's it. It says nothing about what they paid for those picks, how much they staked, or whether the account made a single dollar.

And there's a trivial way to farm it. Buy heavy favorites at $0.95 all day and the market hands you a 90%+ win rate almost automatically — you're "winning" bets the market already priced as near-certain. Whether that makes money is a completely different question. Here's the math.

A 90% win rate that loses money

Take a trader who makes 100 trades of $100 each, always buying favorites at an average of $0.92:

• Each win: $100 at $0.92 buys about 108.7 shares, which pay out $108.70 — a profit of $8.70.
• Each loss: the full $100 is gone.
• 90 wins × $8.70 = +$783. 10 losses × $100 = −$1,000.

Net: −$217, before fees. Ninety percent win rate, red month. And this isn't a weird edge case — the favorite-buyer profile is all over every leaderboard, wearing a beautiful green percentage.

A 40% win rate that prints

Now a trader who hunts cheap entries at an average of $0.25, same 100 trades of $100:

• Each win: $100 at $0.25 buys 400 shares, paying $400 — a profit of $300.
• 40 wins × $300 = +$12,000. 60 losses × $100 = −$6,000.

Net: +$6,000. This wallet loses six rounds out of ten. In a tracker it looks like a coin-flipping degenerate. It's the most profitable profile on this page. (Both examples ignore early exits and fees to keep the arithmetic clean; the direction of the result doesn't change.)

The research says the same thing

This isn't just a toy example. A study of 222 million Polymarket trades, summarized in Federico Glancszpigel's review of the academic literature, found something stranger: traders with above-random forecasting accuracy earned negative returns, while near-random pickers with good execution made money. Being right more often than chance still lost, because accurate forecasters tended to arrive late and pay bad prices. The pick was fine. The price wasn't.

The same review puts fewer than 30% of all traders at a positive return, with the top 1% of accounts capturing 84% of aggregate gains. A separate on-chain analysis by DeFi Oasis found roughly 70% of 1.7 million addresses in the red, with just 668 wallets — the million-dollar-plus club — holding 71% of all realized profit. Another analysis puts the loser share at 84%. The datasets disagree on the decimal; none of them disagree on the direction. Most traders lose, and win rate is not what separates the ones who don't.

What to check instead

• Profit against volume. $9,000 profit on $60,000 traded is an edge. $9,000 on $2 million churned is noise with survivor bias.
• Average entry price. The single best one-number upgrade over win rate. It tells you which of the two profiles above you're looking at.
• Sample size. 400 resolved trades mean something. 15 mean nothing, however green they are.
• Worst stretch. Find the ugliest run in the history. You'll be copying through one eventually — can your bankroll and your nerves take it?
• Category concentration. Plenty of wallets are sharp in sports and coin-flips in crypto. Blended stats hide this completely.
• Realized vs open. A profit figure propped up by unrealized positions can evaporate in an afternoon.

I've watched wallets with 88% win rates grind slowly to zero over two months, exactly like the worked example, while a scruffy 44% wallet quietly tripled. The trackers ranked them in the wrong order the entire time.

Using win rate anyway

Win rate isn't useless — it's fine as a floor, as long as it never travels alone. Pair it with a cap on average entry price and it starts meaning something: "picks well AND doesn't overpay." That pairing is built into btc5min's Market Scanner, which is why the win-rate floor sits next to a max-average-buy-price filter and a minimum-profit filter rather than standing alone. The Wallet Searcher then shows any candidate's per-category P&L over their recent markets, which settles the "sharp where, exactly?" question.

One more trap: the common advice to "only copy wallets above 80%" actively selects for the overpriced-favorite buyer — the −$217 trader from earlier. It also selects for arbitrage bots, whose stats are distorted in their own way; spotting bot wallets covers that. For the wider question of whether any of this makes money, see is copy trading profitable, and for the leaderboard's other traps, how to read the leaderboard.

btc5min is an independent tool and is not affiliated with, endorsed by, or operated by Polymarket. Nothing here is financial advice. Prediction markets are risky, copying another trader does not guarantee profit, and you can lose the money you trade with.