An Application of the No Free Lunch Theorem

[madnak]

But I agree that "averaged over all possible terrains" doesn't really make sense at all. I think there's something fishy going on. How did they come up with their enumeration of terrain types? How did they evaluate each strategy for each terrain type? On what basis did they consider the frequency of the terrain type (or did they at all)? And how did they perform the "averaging?"