What is a good win rate? Why a 90% strategy can still lose money

By John, builder of Wise Apple · Updated August 27, 2026

Ninety percent winners. It is the number in the ad, the number in the Discord, the number your friend will not stop mentioning. It is also the number that, on its own, tells you nothing about whether the account is going up.

Win rate answers one question: how often does a trade close in profit? It does not say how much profit, or how much loss the other trades take, and without those two numbers it is not a performance metric. It is a mood. A strategy can win 90% of the time and lose money steadily. Another can win 35% of the time and compound for a decade. The arithmetic that separates them is one line long, and this guide is about that line.

Expectancy: the number underneath

Expectancy is the average result per trade, and it is the number win rate is hiding:

Take the 90% strategy. Say the average winner is 1 unit and the average loser — the 10% — is 15 units, because the strategy holds losers hoping they turn. Expectancy is (0.9 × 1) − (0.1 × 15) = 0.9 − 1.5 = −0.6 units per trade. Every trade, on average, loses 0.6 units, and the account bleeds while the win-rate counter ticks up.

Now take a 35% strategy where winners average 4 units and losers average 1 unit. Expectancy is (0.35 × 4) − (0.65 × 1) = 1.4 − 0.65 = +0.75 units per trade. It loses two trades in three and makes money doing it. Van Tharp's Trade Your Way to Financial Freedom (1998) is the book that made this arithmetic standard for retail traders, and it has not stopped being true.

The break-even table

For any payoff ratio — average win divided by average loss — there is a win rate at which expectancy is exactly zero: break-even win rate = 1 / (1 + payoff ratio). Above it you make money, below it you lose, before costs.

Read the first row against the 90% strategy above. Its payoff ratio was 1/15 ≈ 0.067, for which the break-even win rate is 1 / 1.067 ≈ 93.7%. Ninety percent was not enough. It was never going to be.

Costs shift every row upward. A round trip that costs a fifth of the average win moves a 50% break-even to something closer to 55%, and the smaller the average win, the larger the shift — which is why the costs guide and this one are the same problem from two sides.

How a high win rate strategy blows up

High win rates are not fake. They are usually real and usually bought with a specific trade-off, and the trade-off is what detonates. Three patterns account for most of the blow-ups:

Selling the tail

Collect a small premium on most days in exchange for a very large loss on rare days — short volatility, far out-of-the-money options, anything that "works until it doesn't". The win rate is excellent because the rare day is rare. The expectancy is often negative because the rare day is enormous, and a backtest that did not contain the rare day cannot tell you so.

Averaging down and grid recovery

Add to losers until the position comes back to break-even, then close it as a small win. The win rate approaches 100%, because every trade is held until it wins — right up to the one that does not come back, at which point the position is many times its original size. This is not a strategy with a high win rate. It is a redefinition of "loss" to mean "position I have not closed yet".

Tiny targets, enormous stops

Take profit at 0.5%, stop at 5%. Most trades hit the small target first. The win rate looks wonderful, the payoff ratio is 0.1, and the break-even win rate is 91%. The strategy is fine until a normal week delivers three stops in a row and erases thirty wins.

So what is a good win rate?

The honest answer is: the one that, with your payoff ratio, gives a positive expectancy after costs, with enough trades that the number is not noise. In practice, the two styles cluster. Trend-following and breakout systems tend to have low win rates and large payoff ratios — they lose often and small and win rarely and big. Mean-reversion systems tend toward high win rates and small payoffs. Neither is better. Each is a different way of being right, and each has a failure mode: trend systems die of a thousand cuts in a range, mean-reversion systems die of one cut in a trend.

One more thing a win rate needs before it means anything: a sample. Sixty-eight percent over 25 trades has a 95% interval of roughly 50% to 86%, which includes losing strategies at most payoff ratios. The sample-size guide has the table.

What to look at instead

1. Expectancy per trade, after costs. The number that decides whether the account goes up.
2. Payoff ratio. Read it next to the win rate; together they are expectancy. Apart, each is half a fact.
3. Profit factor. Gross wins divided by gross losses. Below 1 loses; near 1 is a cost model away from losing.
4. Maximum drawdown and its duration. How deep, and for how long. This is what you will have to live through.
5. The trade distribution. Remove the best 5% of trades. If the strategy stops working, it was three lucky trades.
6. Trade count per regime. Whether any of the above has been measured in the conditions that will end the strategy.

The report-reading guide puts these in the order that catches the most problems soonest. Win rate is on the list. It is not near the top.

How Wise Apple reports it

Wise Apple's report does show win rate — judged on out-of-sample windows, not the training data — but it leads with the trade count, expectancy and drawdown, and it puts the buy-and-hold comparison beside the result so a 90% win rate that trailed simply holding the asset reads as what it is. The 1–100 Score is capped below the midpoint for any configuration that does not beat holding, whatever its hit rate. Every trade in the journal is inspectable candle by candle, which is how you find the fifteen-unit loser hiding behind the ninety percent.