The more strategies you test on historical data and pick the best, the more you are fitting to noise — not finding a real edge. Drag the slider to see how in-sample performance grows while out-of-sample performance collapses.