Author: Tony Werten
Date: 21:36:48 01/24/02
Go up one level in this thread
On January 24, 2002 at 14:05:54, Leen Ammeraal wrote: >In an evaluation function, it makes sense to count >and reward attacks to occupied and free squares. >However, this can be done in several ways. >Currently, I have the following data available >and I assign bonus points for all these three >count results: > >a. counting each square that I attack > at most once, regardless of the number of > attackers and defenders for that square. > >b. counting all my attacks to squares, so this > count is 2 for a square if both my knight and > my bishop (and no other pieces of mine) attack it. > >c. counting all squares that I control in the > sense that I attack it with a given piece > and the opponent defends it only with > more expensive pieces. For example, I regard > a square in my control if I attack it with a pawn > (and possibly other pieces as well), while the > opponent may defend it with as many > non-pawn pieces as he likes. > >If you have a opinion (preferably based on experience >and chess knowledge) about the relative merits of >these counts in an eval function, please indicate >this by giving weight factors. >For example: >c 3, a 1, b 0 (if you regard b as useless). >(I hope your answer will not be a 0, b 0, c 0 -;) Well, almost as usefull: a 1 b 2 c 3 a is better than piecesquare tables I guess (but I'm not sure). Unfortunately it takes more processortime for each step as wel. Find a way to hash the results to keep a little speed. In XiniX I use b, but I use different values for each square as wel. Tony > >Leen
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.