Author: Ron Murawski
Date: 22:47:38 11/13/02
Go up one level in this thread
On November 13, 2002 at 21:51:28, andrew tanner wrote: > > > > I was wondering about the way a typical evaluation function works. I don't think that there exists a "typical" evaluation function. There will be similarities, but it is the differences will determine the engine's presonality. >Is it >not more productive to evaluate the position in terms of sides of the board and >is this done? I ask this because a typical anti-computer chess strategy is to >build up an attack against the castled kingside. The computer typically doesn't >see an early kingside knight sacrifice as a major threat. (see Ed Nemeth's 17 >move mate against shredder 5 ) Ideally as I see it, a computer has the potential >to calculate scores for piece positions on the 2 sides of the board and to use >these scores constructively as the game progresses (with an attack strategy on >one side in mind). So it should be the computer making aggressive deadly >sacrifices early to create direct attacks on the casled king. > A computer never knowingly plays a true sacrifice. All it can do is make the move that will get it the highest score, aka "best move". In order to get an engine to play a knight sacrifice, you must award enough attacking bonuses to outweigh the loss of the knight. The problem is how to do this in a way that the engine doesn't sac away minor pieces needlessly. Some popular approaches are to give bonuses for wrecking the opponent's king shield pawns and awarding bonuses for piece proximity to the enemy king. As a programmer I can tell you that it gets scary knowing you are giving positional bonuses of more than 3 pawns for king attacking possibilities that may not materialize. My own engine attempts to do this and I find that sometimes it works and sometimes it doesn't. > You should have (ideally) the queenside evaluation and the kingside evaluation >for all pieces on those sides of the board. In Capablanca's book "chess >fundamentals" this is the way he approaches chess, and we all know that he >played extremely simple astonishing moves. He also stated that in Grandmaster >play one common way to win was to divert the opponents pieces to one side of the >board, then quickly shift the attack to the other side. I rarely if ever see >computers employing long term strategies such as these unless the position calls >for it. Being able to switch your attack from one side of the board to the other means that you enjoy more mobility than your opponent. I believe that all of the better engines calculate piece mobility into their scoring. Ron
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.