Backtesting a strategy

How many trades do you backtest before trusting a strategy?

A strategy can look excellent over a small sample but perform differently when tested across different market conditions.

I'd like to understand how other traders approach this. Do you test a fixed number of trades, several months of data, or multiple market regimes?

Which statistics matter most to you: win rate, expectancy, maximum drawdown, profit factor, or average risk-to-reward ratio?

I'm especially interested in how you avoid overfitting a strategy to historical data.