Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Draw recognition by eval problems

Author: Andrew Williams

Date: 07:42:45 10/17/01

Go up one level in this thread


On October 17, 2001 at 06:19:46, Rafael Andrist wrote:

>[D]6k1/8/7P/3pK3/8/8/2B5/8 w - - 0 1
>
>As most of the others, my program knows that KBP-K with a rook pawn and bishop
>of wrong colour is drawn if some conditions are met. In the position given above
>black has also pawn and the position remains a theoretical draw. Now, my program
>evaluates it as +3.5 and refuses to capture the black pawn. Of course I could
>add this special case too but there are lots of other combinations of pieces I
>had to add too. Did someone try a more general attempt to solve this problem?
>Would the following rule work in this case? If White is ahead in material and if
>it is a drawn KBP-K position after removing all black pieces, the position is a
>draw.
>

I think you can deal with this by saying in your eval that "White cannot win".
So if you're about to return a score which is > 0 from White's point of view,
you return 0 instead.

>The problem are IMO possible Zugzwang positions which could occur, at least in
>other types of drawn endgames with big material imbalance, e.g. KQ-KP. In this
>case a theoretically drawn position could converted to a lost in some cases if a
>black piece is added.


Andrew



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.