TalkChess.com

[H.G.Muller]

Well, uMax does do IID, but the problem is more that it has no move list to store and sort the moves between generation and execution. The main reason it can be so small is that it immediately performs the moves as the move generator produces them, and all info about the move (from, to, capture square, piece, victim, ep+castling flags) is still available in simple variable. You would either have to store all that in the move list (and access to an array variable A[k] or through a pointer p->A takes 4 characters as opposed to 1 for a simple variable). Or you would have to regenrate most of that info from the stored (from,to) pair only, but that would duplicate a large part of the move generator. And the move geerator is about 60% of uMax already...

So the only thing uMax learns from the IID is the best move, which then become hash move in the next iteration.

So the only thing I have to go on in making the LMR decision is the properties of the moves themselves. The order means nothing. But I don't necessarily consider that a bad thing.