debate about backtesting

How many backtested trades do you consider enough before trusting a strategy with real money? Is 50 trades sufficient, or would you prefer at least 100–200?