The Arithmetic of a 95% Win Rate: Why High-Winrate EAs Still Blow Up

A 95% win rate sounds like the end of the search. It is usually the beginning of the problem. The question that decides everything is not how often you win — it is what the average loss costs relative to the average win.

Expectancy = (win rate x average win) - (loss rate x average loss).

Take an EA that wins 95% of trades, averaging +$2 per win. It loses 5% of trades, averaging -$45 per loss. Expectancy per trade: 0.95 x 2 - 0.05 x 45 = minus 35 cents. This account produces weeks of pleasant green trades and is mathematically guaranteed to bleed. Scale the lots up and the same arithmetic produces the famous "smooth curve, then one red candle ends it" screenshot.

It is what averaging and martingale management create by construction. Every small loss is refused: the position is added to until price comes back, so most sequences end as a small win. The rare sequence where price does not come back collects all the refused losses at once, multiplied by the grown position size. In trade history you can recognize it instantly: average loss 5x-20x the average win, longest losing streak short, deepest loss catastrophic.

1. What is the ratio of average loss to average win? Below 2x is normal for tight-stop systems. Above 5x means the win rate is being bought on credit.

2. What happened — or would have happened — in the worst month of 2020 or 2022? If the test window carefully avoids those, the answer is being avoided too.

Our trend systems win around 40-45% of trades. It looks worse in a screenshot and behaves better in a crash: every trade carries a hard stop, losses stay ordinary, and the average win is larger than the average loss. A win rate you can survive beats a win rate you can frame.

Our products and free risk utilities: https://www.mql5.com/en/users/app.develop.sk/seller