Author: Colin Frayn
Date: 02:54:22 04/03/00
Go up one level in this thread
>Eventually you'll probably want to add some more terms to your quick eval which >will make lazy eval work even better. In particular you have to be extremely careful with checks, and positions where you are in check yourself etc... >This is futility pruning in the q-search. Its a well accepted method, and >saves a bunch of nodes. I found it reduced tactical accuracy noticeably so I left it out of ColChess as I wanted my analysis to be 100% accurate.. >you can do futility pruning in the last few ply of fullwidth but you need to be >a bit careful with it. See Darkthought pages for a good article on it... Heinz, E.A., ICCA Journal, June 1998, pp 75-83. I have it right here :) I read through this one before I tried to implement futility pruning at frontier nodes, and a limited method using reduced width quiescence search (RWQS) futility pruning, which is generally just a method to check that your assumption is OK, and doesn't require _too_ many wild guesses. I discarded the idea after a few weeks though as it messed up my analysis. Cheers, Colin
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.