Author: Antonio Dieguez
Date: 17:14:49 11/05/99
Go up one level in this thread
On November 05, 1999 at 12:03:36, William Bryant wrote: >I would suspect that there may be errors in you implementation. It took several >weeks to debug my null move code. > >You may be doing all of these things, but issue to consider include: > >Make sure you make and unmake null move to set the hashtable signature for >the new side to move, update ep square, etc. > >You must avoid making null moves where not making a move improves you position >or the other side is in check, and if you just made a null move. > >There are a series of test my program does before making a null move. >If I just made a null move -- don't do it again immediately. >If in check, null is out. >If material is low, this sets up for a zugswang(SP-- no dictionary right here >right now)and null should be avoided. > >Also, since the goal of null move is to fail high, avoid doing a null if >there is not chance it will fail high. When you probe the hash table, if the >value is exact, or a upper bound (fail low) score, check the score to see if >there is any point in doing the null move at all. Bob calls this the null move >'trick', which is well documented in the Crafty source code. Look in the file >hash.c which will return a don't do a null move flag. > >William >wbryant@ix.netcom.com Thanks William! Anyway I dont have bugs (ok I think only) in my nullmove implementations, and may be because my depth is from 0 to finaldepth in the search(not from x to 0), and the finaldepth is passed too in the search function, and I dont know, may be that makes more easy the thing who knows. I still dont use hash tables but thanks for the voices(is it well spelled?) Yes, zugzwang is a problem that many pograms have and I have to consider that, there are a lot of things to do. Thanks again, bye.
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.