Author: Robert Hyatt
Date: 18:44:16 11/05/00
Go up one level in this thread
On November 05, 2000 at 20:51:17, Pham Minh Tri wrote: >On November 05, 2000 at 20:24:43, Robert Hyatt wrote: > >>On November 05, 2000 at 17:53:18, Pham Minh Tri wrote: >> >>>Hi, >>>I know that "attacked opponents squares" value is one of the important factor >>>for evaluation function. However, some of free chess engines like Crafty, GNU >>>chess seem not to use it. So I was wondering why. >>>Thanks in advance for any explanation. >>>Pham >> >> >>Crafty does use this, where appropriate. IE king safety. It also uses this >>in Swap() to evaluate captures on a given square... >> >>If you mean the classic definition of "space" then I don't evaluate that in >>the usual sense, no... > >Yes, I mean the definition as the number of "squres are attacked on the >opponent's half of the board" (Advances in Computer Chess 5). The result shows >that it is a very important feature. >I was wondering if: 1) There are some new researches show that it is not >important or it may be replaced by other factors. 2) It is expensive to >implement. 3) Some other reasons? >Thanks, 1. One of the hard things about such evaluation ideas is the answer to the question: "is this a cause of winning the game, or is it an effect (result) of winning the game?" Sometimes space is critical, sometimes it is not. It is possible to recognize space in different ways. IE a rook on the 7th is a form of "space". 2. It isn't terribly expensive to implement. I think gnuchess even has an eval term for hung pieces, which requires this information to recognize. The question is, is the cost offset by the gain? 3. most programs evaluate lots of things, some of which overlap with the concept of 'space' and 'mobility'. It might be that such ideas are good enough...
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.