Author: blass uri
Date: 14:40:32 10/28/99
Go up one level in this thread
On October 28, 1999 at 17:20:26, Alexander Kure wrote: >On October 28, 1999 at 17:06:42, blass uri wrote: > >>On October 28, 1999 at 16:51:56, Alexander Kure wrote: >><snipped> >>>I cannot confirm that commercial programs _do_ learn a lot from crafty's source >>>code. What can be learnt from it? >> >>crafty does some things that some commercial programs start to do only now. >> >>examples: >>1)using nalimov tablebases > > >Using Nalimov tablebases was not totally new. Years before there were Ken >Thompson's Endgame tablebases. I do not know if crafty was the first chess >programm to access egtbs in his search. I know it is using nalimov tablebases for some years and using nalimov tablebases is better than Thompson's tablebases. Thompson's tablebases do not "know" if the position is a draw or a win for the side with less material. > >>2)learning from previous search by not clearing hash tables after every move >> >>crafty has some knoledge in simple endgames that some commercial programs do not >>have >> >>It understand cases when KBPP vs K is a draw. > > >this is a 5 piece ending - no need to know. It understands it also if the opponent has a pawn. Uri
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.