Polymarket Win Rate: Why 90% Wins Can Still Lose Money

A high Polymarket win rate does not guarantee a profit. Buying at $0.95 and winning 90% of the time loses money before fees. Winning 60% at an entry price of $0.55 can produce a positive result. To evaluate a trader, you need the execution price, position size, costs and counting method alongside the percentage of wins.

This guide explains win rate, breakeven and their relationship to P&L with reproducible examples, then gives you a wallet-checking workflow. Every numerical scenario is hypothetical: it illustrates the calculation, not the performance of a real trader or an expected return.

What is win rate, and what counts as a win?

Win rate is the fraction of successful observations in a defined sample: wins ÷ completed observations × 100%. The word “observations” matters. A tracker might count individual fills, closed positions, resolved markets or trades with positive financial P&L. Those definitions answer different questions.

Five purchases of the same outcome are five fills, but they share one market result. Counting each fill separately increases the apparent number of trades without adding the same number of independent decisions. Before comparing two wallets, check the counting unit, date range and whether losses are included.

The calculations below use a simple setup: buy one outcome, hold it through final resolution, receive $1 for a winning share and $0 for a losing share. This mechanism is described in Polymarket’s positions documentation. Markets with a split payout must be evaluated separately using their actual redemption value.

Why a 90% win rate can lose while 60% makes money

Consider two series of 100 purchases, one share per market. Each series uses a constant entry price and holds every share to resolution. Fees and other costs are excluded for now.

In Series A, each win earns $0.05 per share while a loss costs $0.95. One loss wipes out nineteen wins. In Series B, a win earns $0.45 and a loss costs $0.55. Ranking the two series by win percentage alone puts them in the wrong order.

The final row measures profit against the total purchase cost of this hypothetical series. It is not an annual return, a return on the entire account or necessarily the ROI definition used by a particular leaderboard.

How to calculate breakeven win rate

Let p be the purchase price of one share and w the fraction of wins. With a constant price and equal share counts per observation, the result per purchased share before costs is:

w × (1 − p) − (1 − w) × p = w − p.

Breakeven occurs when w = p. A $0.65 entry requires a 65% win rate for a zero result before costs. An $0.80 entry requires 80%. Profit requires beating that threshold. For the underlying relationship between price and probability, see how to read Polymarket odds.

The broader formula uses average profit on a winning observation G and average absolute loss on a losing observation L: breakeven win rate = L ÷ (G + L). G and L must refer to comparable completed positions and treat costs consistently. This is the link between win rate and risk–reward: how much each outcome earns or loses matters as much as how often it occurs.

Costs raise the threshold

If total costs were a hypothetical $0.01 per purchased share, a $0.65 entry would require 66% wins to break even. One cent is an illustrative assumption, not a Polymarket fee quote. Actual fees depend on the market and execution; check the current official fee documentation.

When using actual average fill prices, slippage is already reflected in those prices. Do not subtract it a second time. When starting from a displayed quote, first account for the difference between that quote and the executed price. Reconcile fees, rebates and other cash flows separately. The fees, deposits and withdrawals guide provides a starting point for reviewing costs.

Why average entry and win rate are not enough when sizes differ

“Win rate minus average entry” describes the result per share when every observation contains the same number of shares and is held to resolution. With unequal sizes, an ordinary position-count win rate can conceal a loss.

Suppose nine winning positions contain 10 shares each, while one losing position contains 100 shares. Every share cost $0.60. The position-count win rate is 90%. Yet the purchase cost is 190 × $0.60 = $114, the payout is 90 × $1 = $90, and P&L is −$24 before costs. The biggest position lost.

For unequal sizes, calculate the result of each purchase. Let qᵢ be its share count, pᵢ its actual entry price and yᵢ the payout per share. For purchases held through resolution:

P&L before costs = Σ qᵢ × (yᵢ − pᵢ).

With $1/$0 payouts, the comparable pair is the fraction of winning shares and the average entry price, both weighted by share count. A simple average across positions misses where the exposure sits. The position-sizing guide discusses the relationship between size and risk.

Early exits need a different calculation

When a trader sells before resolution, financial P&L depends on the exit price, not just the final outcome. Buying at $0.45 and selling at $0.60 earns $0.15 per share before costs. The closed trade remains profitable even if the purchased outcome later loses.

For a simple fully closed position, use share count × (sale price − purchase price) − costs. Multiple entries and partial exits require a consistent cost-basis method. A win rate based on market outcomes must not be mixed with the percentage of financially profitable exits.

Keep open positions separate from realized results. Marked value changes with the market price and can differ from the amount obtainable in an actual sale. Deposits, withdrawals and token transfers are not themselves trading gains or losses. Additional rewards, where applicable, should not be presented as profits from predicting outcomes.

Check sample quality, concentration and subsequent performance

A 70% win rate might mean seven wins out of ten or 700 out of 1,000. The percentage is identical; the amount of evidence is not. More history helps, but there is no fixed trade count after which skill is automatically proven.

• Dependence. Multiple fills in one market or positions tied to related events are not independent observations.
• Profit concentration. Check whether the result remains positive without the single most profitable market. Use this as a robustness diagnostic, not as permission to remove inconvenient trades from the reported record.
• Selection after the fact. Screening many wallets can uncover an impressive record by chance. Once selection rules are fixed, evaluate a separate subsequent period.
• Coverage. A limited sample, missing losses or starting inventory with an unknown cost basis can distort the result.
• Changing conditions. Fees, size, liquidity and the markets traded can change even when the approach appears unchanged.

The gap between historical win rate and entry price helps explain a past series. It does not automatically estimate the true probability of the next outcome. Keep that distinction in mind when reading about Polymarket edge. An unconditional historical UP frequency is not a ready-made probability for a particular entry either.

A practical Polymarket wallet checklist

1. Choose an address, period and market type. Start with the wallet analytics dashboard or Up/Down trader leaderboard. Compare matching periods and note coverage limitations.
2. Identify the counting rule. A winning market outcome, an individual fill and a profitable closed position are different units.
3. Review prices and sizes. Examine actual entries and the distribution of exposure. If costs or starting inventory are unknown, keep that limitation explicit.
4. Separate the components. Realized profit, open positions and additional payments should not become one unexplained total.
5. Inspect the history. Find the largest gains and losses, clusters of activity and markets responsible for most of the result. The wallet P&L and history guide explains how to read the available metrics.
6. Evaluate what happens next. Record the selection date and criteria before observing subsequent trades. The record that made a wallet rank highly is not the same as evidence after selection.

A useful question is: “What result did this trader achieve after costs, at which prices and sizes, and does it persist on new data?” Win percentage supplies only part of the answer. This guide explains calculations and interpretation; it does not recommend a particular trade.

Sources and calculation method

Position and fee mechanics were checked on 13 October 2026 against Positions & Tokens and Fees. The tables and scenarios were calculated for this guide. They contain no real-wallet observations, estimated future probabilities or claimed strategy returns.