Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Why this move is so difficult for engines ?

Author: martin fierz

Date: 15:17:34 10/17/02

Go up one level in this thread


On October 17, 2002 at 13:19:09, Dieter Buerssner wrote:

>On October 17, 2002 at 04:51:28, Christophe Drieu wrote:
>
>>[D] 8/4k3/4p3/pp5p/6p1/2P2rP1/PPK2R1P/8 w - - 0 1
>>
>>Rxf3 !
>
>I think, this is not easy at all for engines. Sure - some code that takes into
>account that far away passed pawns (or candidates) can't be stopped by a K while
>some connected passed pawns can, would help here. But, is such code correct?
>Can't it also lead the engine into some trap, and make it thereby less objective
>(for example for long analysis)?
>
>If I look at the position after 1. Rxf3 gxf3 2. Kd3 e5
>
>Now a5, h3 or Ke3 win. But b3 (with the obvious idea to create the passer on
>Q-side, and later one on K-side) only seems to draw. I fear, it is almost
>impossible without any concrete (and in this case probably rather deep)
>calculation, for an engine, to give the position after b3 a drawing score, and
>the position after h3 a winning score (for white). Wouldn't many intermediate
>strength chess player move b3 here?

absolutely. on general grounds, b3 is also a very obvious move. this position is
clearly about calculation and not only about knowledge. i think any program that
evaluates this pawn endgame as totally winning in it's evaluation function will
fail in similar cases where black is just fast enough with his counterplay.

a reason some engines may have trouble with this is that they cannot reach a
good enough search depth for the pawn endgame to see it's winning from the
starting position, because they also have to search the rook endgame.

aloha
  martin



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.