Backtesting Before Increasing Risk

I believe traders should collect enough data before changing their risk or strategy. A short winning streak does not necessarily mean the system has improved. How many backtested trades do you consider enough?