Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Crafty score in this endgame

Author: Uri Blass

Date: 07:14:11 03/10/01

Go up one level in this thread


On March 10, 2001 at 09:38:13, Michael Fuhrmann wrote:

>On March 10, 2001 at 01:58:59, Uri Blass wrote:
>
>>On March 10, 2001 at 01:38:25, Pete Galati wrote:
>>
>>>On March 10, 2001 at 00:26:57, Michael Fuhrmann wrote:
>>>
>>>>[D] 8/8/7P/5kP1/5P2/r7/5K2/8 b - - 0 114
>>>>
>>>>
>>>>Crafty scores this as -3.26 for white on my computer, yet it seems to be a draw.
>>>>I don't have all the 5-piece tbs. Is that the problem?
>>>
>>>I don't know if that's a problem.  I have some of the 5 piece tablebases,
>>>whatever came with the Junior CD.
>>>
>>>It doesn't appear to be a draw to me, White looks like it's in a lot of trouble
>>>and it's only hope is for Black to blunder.  I somehow can't imagine White ever
>>>promoting any of those pawn considering that the King can't get up there to help
>>>them promote.  To ME at least, all hope looks lost for White, but I could be
>>>wrong.
>>
>>The scores of Crafty suggest that it is a draw.
>>White cannot promote but black also cannot make progress.
>>3.26 pawns for black at depth 9-13 and
>>3.24 pawns for black at depth 14-17 suggest that black cannot make progress.
>>
>>The most important tablebases for this endgame is the KRKPP tablebases but
>>seeing the draw is not an easy problem for computers because black can play a
>>lot of moves when both sides do no progress and I doubt if they can see it even
>>with the KRKPP tablebases.
>>
>>Uri
>
>Actually, when this position came up in a game between crafty (black) and little
>goliath, little goliath's score was 0.0 while crafty kept showing -3 or
>thereabouts for about 20 futile moves as it looked for a way to win. The fact
>that lg was able to see the draw while crafty's score was so big is what struck
>me. And LG was using no tablebases.

I guess that lgoliath's value of pawn is too high and this is the reason that it
could see the draw.

I guess that reducing the value of pawn could help lgoliath to be a better
program but it is not going to help the program to be better in test suites and
I guess that the programmer used test suites to decided about the evaluation.

Uri



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.