Author: Stuart Cracraft
Date: 07:17:09 09/22/04
Go up one level in this thread
On September 22, 2004 at 01:56:44, Rick Bischoff wrote: >Hello, > >I am using a variant of Ed Schroeder's move ordering technique described on his >webpage-- part of this is move ordering based on a delta from a piece table >(e.g., Moves like Nf3-g1 will get a negative score and be searched close to >last) . I have this all working pretty good and I have it implemented for 3 >different phases. However, I am *CLUELESS* as to what values I should put into >the queen table. What makes a square good for a queen? > >Also, I implemented a working copy of SEE now and have been experimenting with >it-- is it worth the time it takes to execute a SEE to evaluate non-capturing >moves to see if they are hanging? Hi -- I've been using this method for years -- I use the pc/sq tables that are used with material at end points for an evaluation also as ways to do move ordering (along with history heuristic, SEE values, and bonuses for promotion, castling, etc.) For the queen pc/sq table everything is set to 0. Why? I don't want the queen hanging out. However, I do modify squares in the queen pc/sq according to a formula for their distance from the enemy king. Since this is done at the root, error creeps in the further out the evaluation end point is. To counteract this I include a mate threat that is an exponential that goes up with the number of pieces within 2 squares of the enemy king that are occupied by my pieces. On this last feature, when I introduced it I wasn't sure what to expect. I knew it would slow down my evaluator. However, it produced an equal result with my chosen test (Win-at-Chess) as without it -- so in this circumstance it could be left in. Stuart
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.