Help!hgm wrote:All this talk leads nowhere. Let us consider a real example. Suppose I have an LMR tree where each node has 10 moves, and the first 2 are searched at full depth, and the remaining 8 are reduced 1 ply. How would you take the p_i and w_i?
I fully agree to the quoted part that I emphasized above. But instead of continuing to discuss on the level of (old or new) examples, would all of you, HGM, Michel, and Daniel, please give a brief but sufficiently precise description of the method itself (algorithm, formula, ...) that you prefer for estimating perft(X), e.g. perft(13), without repeating your arguments or counter-arguments? I am running into big trouble to follow your discussion simply for the reason that the biggest part of it is repetition of arguments why A misunderstood B and why C is right and D is wrong
Please, to all of you three, if you try to do that, just limit yourself to describing the method, without any stuff like "which is what I already wrote 27 posts ago while all others consistently ignored it"
Sven