This thread is one major backtesting fantasy. The first page is basically about curvefitting the indicator to "maximize" performance. And this is a fine example why this is relatively useless.