Author: Bas Hamstra
Date: 05:05:24 08/17/99
Go up one level in this thread
I'm glad it doesn't work. I hoped so :) In fact I don't like the whole idea of razoring. Pruning because an Eval looks bad? Very strange. I could imagine it would work if you use reasonable limits. For example if at the last ply Eval is QUEENVALUE lower than BestLastPlyEval it is probably ok to to prune. Is that how it is done? BTW would it not be better to simply use Alpha and Beta? IE: if Eval > Beta+KNIGHTVALUE prune (last ply). Idea: Material is good AND I'm to move, so it would probably give a beta cutoff anyway. And if it's NOT so, the worst case is that some tactics are found a ply later. Has anyone tried this? Regards, Bas Hamstra. On August 17, 1999 at 05:32:39, David Blackman wrote: >On August 16, 1999 at 17:47:10, Dave Gomboc wrote: > >>On August 16, 1999 at 10:22:14, Bas Hamstra wrote: >> >>>What good is razoring? How safe is it? Pros en cons? >> >>It prunes at depths 1 to n-1 instead of from depths 2 to n, so it speeds things >>up a lot. It isn't a safe heuristic like alpha-beta. It will blow up in >>zugzwang positions, just like null-move. >> >>Dave > >My impression is it can blow up in all sorts of positions, and usually does it >even worse than null-move. It does prune away a lot of branches, which helps you >get deeper. But too many of the branches it prunes seem to be important ones. >Use with extreme caution, and only near the leaves of the tree. > >My program used to do razoring at most depths, but it made too many tactical >errors. Then it only did razoring when a few complicated heuristics said it >would probably be safe. This helped on some tactical test suites but apparently >not in real games. Now it doesn't do razoring at all, and it is stronger, but >still not as strong as i'd like.
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.