Timeframe changes strategy behavior
The same algorithm on 5m, 15m, 30m, and 1h can behave very differently: signal frequency, average trade duration, drawdown, and noise sensitivity all change.
A lower timeframe usually creates more trades and higher execution-quality requirements. A higher timeframe may create fewer entries, but can sometimes filter random movement better.
Test different historical periods
Optimization depends on the selected historical window. Even moving the start or end date by 1-2 days can materially change the result, especially if that segment contained sharp moves.
Do not rely on one run. Compare several windows: a recent short segment, a longer period, and a period where the market behaved differently.
Look for robustness, not a beautiful accident
The best result at one point is not always the best live candidate. Prefer settings that keep acceptable drawdown, enough trades, and understandable logic across several checks.
A practical workflow is to try more timeframe, period, risk, and strategy-parameter variants, then keep the result that is not just beautiful, but explainable and stable.