Backtesting2026-06-11

Why a high win rate can still lose money

TL;DR

Yes, a strategy can win 90% of the time and lose money, because win rate only says how often you are right while expectancy says how much being right is worth. Ninety small wins against ten oversized losses nets out negative. The number that decides is expectancy after costs, and your own trade log already contains it.

Can a strategy with a 90% win rate lose money?

Easily, and the arithmetic is short. Say the strategy wins 90% of the time at +0.2R and loses 10% of the time at 3R. The wins contribute 0.18R per trade, the losses take 0.30R, and the system bleeds 0.12R per trade while winning nine times out of ten. Run it a few hundred trades and the equity curve grinds down with a smile on its face.

Nothing about that example is exotic. Selling far out of the money options, holding losers past the stop, scalping with a wide emergency exit, all of them buy a high win rate by renting it from the left tail.

The market pays you in R, not in percent right.

What is expectancy and why does it beat win rate?

Expectancy is what a trade is worth on average once wins and losses are netted: win rate times average win, minus loss rate times average loss. It is the only headline number that contains both halves of the record, which is why Tharp built his whole evaluation framework on it and on R multiples rather than on hit rate.

Two strategies can share a 60% win rate and sit on opposite sides of zero. The win rate cannot see the difference. Expectancy is the difference.

Why do high win rate systems hide their risk so well?

Because the feedback is almost always pleasant. A system that wins 90% of the time hands you week after week of small green, the journal looks like skill, confidence compounds faster than the account, and the tail event that defines the whole system might not arrive in the first hundred trades. The sample feels large because the days felt full. It is not large where it counts, in the left tail, and thin samples flatter exactly this profile.

A 90% win rate is the most expensive feeling in trading.

Where does risk of ruin come into it?

The tail does not just drain the average, it concentrates the damage into single events, and that shape drives risk of ruin harder than win rate ever can. A high win rate system with a 3R tail carries more ruin at the same size than a 45% system that never loses past 1R, because the floor only needs to be touched once.

One caveat before the win rate gets dismissed entirely: psychologically it is not nothing. A 40% system with positive expectancy is mathematically fine and emotionally brutal, six losses in a row arrive on schedule, and plenty of traders abandon a working system mid streak. The honest read is that win rate prices the experience of trading a system, and expectancy prices the system.

How do you check your own numbers?

Pull the trade log and read three things: expectancy per trade after costs, the size of the average loss against the average win, and the worst 5% of trades. Quantprove reports all three from one upload, as EV per trade, the asymmetry split, and CVaR 95, and the Edge Score weighs the tail directly through its downside bucket.

ProfileWhat the record showsVerdict
High win rate, small wins, deep tail+0.2R typical win, occasional 3RNegative expectancy in disguise
High win rate, controlled tailLosses capped near 1RReal edge, comfortable to trade
Low win rate, big winnersMany 1R, occasional +4RReal edge, brutal to sit through
Low win rate, small winnersNothing covers the lossesNo edge, no disguise

The purest form of the high win rate trap has a name, and traders keep rediscovering it with fresh confidence: martingale. That one gets its own piece. Before then, run your own log and read which row of the table you are in.

Frequently asked questions

It is pleasant, and by itself it proves nothing. A high win rate with capped losses is a real, comfortable edge. The same win rate funded by oversized losses is a losing system with good manners. Expectancy after costs is the number that separates the two.
Everything from 35% to 90%, because profitability does not live in the win rate. Trend followers profit while losing most trades; mean reversion systems win most trades with thinner margins. The pairing of win rate with payoff shape decides, never the rate alone.
Positive after real costs is the bar that matters, and in Quantprove's Backtest scoring the EV per trade input reaches its cap at 0.50R. A small positive expectancy traded frequently can outrun a large one traded rarely, which is what the Edge Velocity metric reads.
More than feels necessary, because the deep tail shows up rarely by definition. Quantprove reaches full statistical weight at 500 trades. A 90% win rate over 60 trades with no tail event on record is an incomplete story, not a safe one.

References

See how many trades your strategy has earned.

Upload your trade log and read your Edge Score with its sample-size adjustment. Free to start.

Start for freeHow it works

No credit card required·Swiss Made