TalkChess.com

[H.G.Muller]

I agree with that, and I will certainly perform such tests in the future, when I am set up better to do this. (I would have to obtain a sufficiently large number of suitable opponents, for instance.)

Nevertheless, comparing two differently shaped search trees against each other at equal (average) number of nodes, using the same eval in both, still seems a meaningful test for how well a pruning method is able to discriminate between useful and useless parts of the tree. Even if you cannot directly translate it into an ELO number.

It still would help, though, to know what is the best one can expect from a technique like LMR. So I would still be curious to how many ELO points others have been able to gain through LMR, and how advanced they had to make it to get there.