Probably lucky outliers. Any data mining driven approach that shows good returns is arbitraged out of existence almost immediately, unless it's niche and thus

Probably lucky outliers. Any data mining driven approach that shows good returns is arbitraged out of existence almost immediately, unless it's niche and thus unscalable to a larger fund. All the blackbox systems end up doing pretty much the same thing and getting low returns. The markets are simply too dynamic for a pure data and algorithms system to work. One needs to use theory to win, and the guy picking the theories in the program is exactly the same thing as a manager in the first place. The only way to get rich from a naive blackbox is to twist the risk knob up to 11, which always goes badly eventually.