Backtesting Before Going Live

How much backtesting do you think a trader should do before risking real money on a strategy? Is there a point where enough backtesting becomes unnecessary and live execution becomes more important?