Author: Bas Hamstra
Date: 11:02:23 08/14/99
Go up one level in this thread
>would it be accurate to say recursive nullmove is basically letting one side >move alot of times? i know any forward pruning involves risk. could you give a >sample position where recursive nullmove works poorly(but not because of >zugzwang)? thanks Think so: in *every* position in stead of playing a legal move you pass. Ie the side to move does nothing. Now you search just as if the side on move has made a normal move: turn goes to opponent. But in stead of searching to the normal Depth, you search to Depth-R. Idea is that *normally* with two moves in a row the opponent gets a good position. However if it *still* can get no good position, then his position is so rotten, you prune that part of the tree away. Recursive because in a nullmove search (to Depth-R) you again do Nullmoves and search there to Depht-R-R. And so on. Recursively. Only restriction is that you don't let occur 2 Nulls right after each other. It's nonsense to let this happen: white passes, black passes, white passes. So white passes, black must play a real move, white passes etc. You prune a lot of nodes away. And sometimes it trhows away too much, and oversees deep winning combinations, because nullmove concluded the position was rotten. In theory it's possible that (recursive) nullmove does not see a combination, even with infinite searchtime. With non recursive nullmove (not doing nullmoves in a nullmove search) this is not so, because the brute force part of the search picks it up sooner or later. Regards, Bas Hamstra.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.