After three wins, a setup feels infallible and size creeps up. After three losses, it feels broken and gets abandoned right before it works. Both reactions ignore the sample-size needed to judge anything.
A fixed evaluation window (every 30 or 50 trades) and fixed sizing rules keep recent results from driving decisions.
Example: a strategy with a real 45% win rate loses five in a row, which happens about 5% of the time by chance. The trader quits it and switches to something untested.
Related: sample-size, gamblers-fallacy, overconfidence, trading-plan