Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Why can't Rebel solve this simple ending ??

Author: Bela Andrew Evans

Date: 12:12:31 12/30/97

Go up one level in this thread


On December 30, 1997 at 01:14:48, Jouni Uski wrote:
>This 5 piece ending is from CSS 6/1997:
>
>W: Kg3 Qe5 pf4
>B: Kd8 Qd7 to move
>
>According to writer - Qe7 is only drawing move. Rebel 9 don't find
>it in one hour, when other programs play it in 0..2 seconds! Why?
>Tablebase users: please test if Qe7 really is only drawing.

I tested the position with R8, and it seems that R8 evaluates the
king and pawn endings arising after 1...Qe7 2. Qxe7 or 2. Kg4 Qxe5 as
much better for white, not realizing these positions are easy draws.
Why?  If you set up those k&p endings directly, R8 will quickly
figure out they are draws -- but it takes upwards of 26 ply for
the 1...Qe7 2. Qxe7 line, and 21 ply for the 2. Kg4 Qxe5 line.  In
the original position, with queens on the board, R8 would have to
search something like 28 ply to see the draw -- but with two queens
able to move about on an open board, 28 ply aint gonna happen for a
few years.

This analysis would seem to indicate that Rebel would be prone to
mistakenly simplifying favorable positions into drawn k&p endings
because of inadequate search extensions or lack of k&p ending
knowledge.



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.