[hamstergene]

I can't help but notice how these studies always ignore that one's presence on the market changes the market. Past data without your bids is a slightly different data set than the same data with your bids. Every trade you would have got would mean someone didn't get it and instead tried some other trade, which isn't in the data set that is being trained on. Or maybe it would be your strategy that would fail a desired trade: market isn't a grocery store, bids are not guaranteed to be fulfilled. The difference is more and more the bigger amounts of money one flexes.

Perhaps this effect is ignorable if one trades insignificant volumes, but then, one can't get rich on trading insignificant volumes, so what's the point.