Backtesting is not only a way to check signal accuracy, but also a tool to continually improvement. Data can be used for self adjustment as each stock have a individual behavior. Accuracy and return is not the same. A stock signal may provide low statistical accuracy, but the signals may prove high return in average. Sharp market dips or fundamental changes (earning reports, news updates etc), may force a short term reaction and disturb the signals (Signals may jump quickly from buy to sell, just to return to buy).