Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Why a Pawn Center?

Author: KarinsDad

Date: 10:43:42 01/20/99

Go up one level in this thread


On January 20, 1999 at 12:58:08, Laurence Chen wrote:

>This is one of the elements in chess which is very difficult to understand, it
>still puzzles me a lot. The problem is how do you know that you got control of
>center? Is there a way to measure this, if so, how? This is an interesting
>question because chess engines nowadays are "taught" or at least programmed to
>play for the control of the center, so how does one measure this control? Chess
>textbooks tell us that it is imperative to play to control the center, and show
>examples of what happens when one side totally ignores center control. I would
>be interested in the point of view to this question from all levels of
>chessplayers, from novice to master, and chess programmers, too.

My take on this is that for a chess computer, understanding true control of the
center is difficult.

In the program I am writing, I have a square control portion of the evaluation
function. Effectively, the side with the least powerful pieces attacking or
protecting a square is in control of that square (this is a simplistic
explanation of it). This allows me to give a delta score based on which squares
are controlled by which side. For example, it is usually more important to
control d4, d5, e4, and e5 than it is to control a1, a8, h1, and h8.

The difficulty for my program comes in via hidden pieces. For example, a bishop
on b7 assists black in controlling the center, however, if black also has a
knight on c6, then my program does not detect that the bishop is doing that
task. A human would spot it instantly. Hence, the need for a search engine to
see whether moving the knight in given positions re-establishes control for
black over two of those center squares.

The search engine also enables the program to know whether a given move (usually
a pawn push or possibly even a minor piece sac) will bust up one sides control
of the center.

However, again the same problem exists. A human can quickly detect that a pawn
push may in the future bust up the center. For one reason or another, that move
will not currently do the job. Other moves must be made first. However, most
programs do not currently have enough understanding of the game to realize this.
Hence, if proper time for the pawn push is outside of the program's search event
horizon, it will not detect the move until it is possibly too late.

Another advantage that humans have is that they can look at the last few moves
that their opponent made and try to predict why they made them. From this, a
human can detect things such as "My opponent is making moves which will
eventually allow him to bust up my control of the center by pushing that pawn.".
From this, the human can make an evaluation as to the threat level and take
countering actions as necessary. The computer can only do this once the threat
enters it's event horizon (granted, humans have event horizons as well and can
also fall into an opponent's plan too late to realize what was really going on).

Programs can try to predict an opponent's plan and hence possible next moves
based on their last few moves to use in move ordering (and in fact my program
will one day do that), however, again it is easier for a human to understand
this concept than to implement it in a program. For example, the moves 5. cd cd
in a lot of positions is merely black's response to white taking his piece: take
it back. There is often no additional threat associated with this and it does
not necessarily indicate a follow up plan. However, a program would have a more
difficult time determining that cd by black was merely recovering the piece once
plan prediction is added to the program.

KarinsDad



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.