Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Fritz 6 Endgame Weirdness (Crafty is Immune)

Author: Len Eisner

Date: 19:34:36 12/21/99

Go up one level in this thread


On December 21, 1999 at 19:09:04, blass uri wrote:

>On December 21, 1999 at 18:32:11, Len Eisner wrote:
>
>>On December 21, 1999 at 09:11:08, James T. Walker wrote:
>>
>>>On December 20, 1999 at 23:46:37, blass uri wrote:
>>>
>>>>On December 20, 1999 at 21:48:29, Len Eisner wrote:
>>>><snipped>
>>>>>In any case, the position I mentioned should be covered by the KPP vs KP 5 >piece tablebase which I do have loaded.  So I don't know why Fritz 6 (and >H7.32 for that matter) plays f7 since it is obviously not the fastest winning >line.
>>>>
>>>>f7 is the fastest winning line that fritz6 and hiarcs7.32 know by tablebases
>>>>of H732.
>>>>
>>>>They know nothing about kqp vs kp so the fact that there is a faster winning
>>>>line(leading to KQP vs KP) is not relevant
>>>>
>>>>
>>>>Uri
>>>
>>>Hello Uri,
>>>I believe you are right.  He should have the KQPKP tablebase to make this
>>>endgame work properly.  It may even be counter productive to have any "Pawn"
>>>tablebases without the proper "Promotion" tablebases to follow up.  For the
>>>KPPKP tablebase you should have the KQPKP and KPPKQ at the very minimum.  These
>>>endings are usually easy wins for the side with the Queen but that does not mean
>>>that they are not required for the programs to transition into them correctly
>>>from the KPPKP ending.
>>>Jim Walker
>>
>>
>>Hopefully, the KQPKP and KPPKQ will be included on the Chessbase Endgame Turbo
>>CD. The files are too large for me to download.
>>
>>I still don't understand why KppKp does not know the shortest winning line.  Why
>>have it if it is not usable without the other tablebases?  I assumed it would
>>know the best pawn to promote, and after that, the program would be on its own
>>unless you had the other tablebases.
>
>The reason is that the tablebases do not include moves but only distance to mate
>information for every position and the engine does not know to promote
>the pawn because the distance to mate is not known after promoting the pawn and
>is known after other moves.
>
>it may prefer always mate in 20 by not moving the pawn and not shorter mate by
>promoting because the tablebases do not have the number of moves of the shorter
>mate.
>
>The right way to use tablebases(if you have only part of them) is to stop using
>them when the distance to mate is decreased but Hiarcs732 and Fritz6 do not know
>it(crafty fixed this problem).
>
>Uri

You're right.  I tried the position with Crafty 16.18, and it correctly played
g8=Q with mate in 8.

Again, the position is 8/6P1/5P2/1k6/8/p7/8/6K1 w - - 0 1

Len



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.