Why High-Accuracy Trading Strategies Can Still Lose

A high-accuracy or high-win-rate strategy can lose money when its average loss, tail loss, or trading cost is too large relative to its typical gain. Accuracy describes frequency, not economic value.

The payoff structure

Assume a strategy wins 85% of trades, earns 0.5R on the average winner, and loses 4R on the average loser. Gross expectancy = (0.85 × 0.5R) − (0.15 × 4R) = −0.175R The strategy can produce long runs of small gains while remaining unprofitable over a sufficiently representative sample.

Where hidden tail risk appears

High-win-rate behavior can be created by:

• taking small profits and using distant stops;
• averaging into losing positions;
• selling options or volatility without sufficient tail protection;
• grid or martingale sizing;
• leaving losses open while recording closed winners;
• operating only in a market regime that later changes;
• ignoring gaps, slippage, financing, or assignment risk.

Not every high-win-rate strategy uses these structures. The point is to investigate how losses are generated rather than infer safety from accuracy.

Warning signs

• one loss erases many prior wins;
• stop distance grows after entry;
• the largest loss is excluded as “exceptional”;
• unrealized losses are omitted from performance;
• position size rises during adverse movement;
• returns depend on stable correlation or volatility;
• no stress test covers gaps or illiquidity.

Better evaluation

Report expectancy, maximum adverse excursion, largest realized and hypothetical gap loss, drawdown, recovery time, exposure concentration, and costs. Use stress scenarios that are more severe than the observed history.

Prop-account implications

A rare loss can breach a hard account boundary before the strategy’s long-run frequency becomes relevant. Cap idea-level exposure, test the loss against the remaining buffer, and avoid treating recent accuracy as additional risk capacity.