Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: With tablebases play programs stronger!

Author: Russell Reagan

Date: 15:42:07 04/24/02

Go up one level in this thread


On April 24, 2002 at 18:37:40, Dann Corbit wrote:

>On April 24, 2002 at 18:18:08, Eduard Nemeth wrote:
>
>>I played on the fritzserver a game comp.- comp. and in this position
>>I seen, that the Tablebases here shaw only draw-move for white with Bd7!
>>
>>I'm not sure that all Grandmasters this move woult plays?
>>
>>(german: Ich bin mir nicht sicher dass alle Grossmeister disesen Zug sehen
>>wuerden)
>>
>>[D]8/5k2/2B5/8/6pK/8/8/6r1 w - - 0 0
>>
>>TBs says: Bd7 is only dwaw move Kg5 is a mate in 37 moves :)
>
>Not convincing.  Play 1000 games of some program with tablebase files against
>the same program without.  The tablebase provides perfect accuracy, but slows
>down the search.  Les Fernandez did a long experiment and found that the result
>was a push -- neither stronger nor weaker.  Of course that was for the program
>crafty only, so your actual mileage may vary.
>
>On the other hand, tablebase files do make programs play much prettier chess.
>And they may help weak programs a lot (if they can make it that far).

I've never understood why this would slow down the search. Either you have few
enough pieces left to do a TB probe, in which case you do so and retrive a
perfect evaluation and your search is over, or you don't have few enough pieces
so you don't probe. It seems the only "overhead" is a single test to see if
there are few enough pieces left on the board. What am I missing here?

Russell



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.