How do I avoid overfitting my trading strategy?

Overfitting is tuning until the strategy memorizes the past's noise. Defenses, in order of value: (1) fewer parameters — every knob is a chance to fit noise; two or three beats ten; (2) test out-of-sample — develop on one period, validate on another you never touched during tuning; (3) demand robustness — if RSI-30 works but RSI-28 and RSI-32 fail, you found noise, not signal; (4) prefer round numbers and conventional values over precisely optimized ones (14, 30, 70 — not 13, 27.5, 71); (5) count your attempts — after fifty variations, your 'best' result is partly selection luck. The strategy that survives these tests will look less impressive in-sample and more honest out of it. That's the trade you want.