A high-win-rate strategy can lose money when its losing trades are larger than its winners or when trading costs consume its gross gains. Win rate measures how often a trade wins, not how much the entire sequence earns.
A numerical example
Consider ten hypothetical trades. Eight earn $10 each, while two lose $60 each. The strategy wins 80% of the time but earns $80 and loses $120, leaving a $40 loss before costs.
In this example, average gross profit per trade is 0.8 × $10 minus 0.2 × $60, which equals negative $4. Costs would make the net result lower. This is arithmetic, not a forecast for any strategy.
Sizing and costs change the outcome
Real trades may use different position sizes. A small winner and a large loser do not offset simply because each counts as one trade. Fees, funding and slippage also change the result.
AlphaNet’s discussion of AI-agent experiments highlights the same distinction: a model’s trade-count success rate can diverge from its account-level profitability.
The more informative question is whether the full distribution of outcomes remains positive after costs. Average wins and losses, tail losses and position sizes explain the economics that a win-rate headline leaves out.