Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Neverending story with incomplete tablebases

Author: Uri Blass

Date: 22:44:13 08/11/03

Go up one level in this thread


On August 12, 2003 at 01:15:47, Anson T J wrote:

>On August 11, 2003 at 03:04:55, Uri Blass wrote:
>
>>On August 11, 2003 at 01:36:45, Harald Faber wrote:
>>
>>>On August 10, 2003 at 13:57:00, Amir Ban wrote:
>>>
>>>>On August 09, 2003 at 19:11:16, Harald Faber wrote:
>>>>
>>>>>On August 09, 2003 at 05:55:43, Tony Hedlund wrote:
>>>>>
>>>>>>How can J8 allow game 14 to end as a draw?
>>>>>
>>>>>
>>>>>This is no serious question, isn't it?
>>>>>It has been posted and mentioned approx. 1,000,000 times. It is the problem of
>>>>>incomplete tablebases. AFAIK you/SSDF uses the "Fritz Endgame Turbo". Based on
>>>>>the Chessbase marketing it consists of 4 CD-roms with "the most important
>>>>>endgame tablebases". Based on what you saw in this game, what many many others -
>>>>>include me - have seen in similar positions, it is crap.
>>>>>Let me ask you a question. If your answer is yes, do I win a price?
>>>>>Here is the US-$1,000,000 question:
>>>>>Are you missing at least one of these files in your tablebase-directory?
>>>>>
>>>>>kqbkp.nbb.emd
>>>>>kqbkp.nbw.emd
>>>>>kqbkn.nbb.emd
>>>>>kqbkn.nbw.emd
>>>>>kqbkb.nbb.emd
>>>>>kqbkb.nbw.emd
>>>>>kqbkr.nbb.emd
>>>>>kqbkr.nbw.emd
>>>>>kqbkq.nbb.emd
>>>>>kqbkq.nbw.emd
>>>>>
>>>>>
>>>>>If you had them all(!), Junior 8 would have mated.
>>>>>
>>>>>Don't you think it is time to consider
>>>>>a) adding the missing 5-man files or
>>>>>b) kick the existing 5-man?
>>>>>
>>>>>
>>>>
>>>>I don't understand this explanation. Are you able to recreate this by omitting
>>>>some 5-man files ?
>>>
>>>
>>>Didn't try it. Have removed the files above from the tb path now but also cannot
>>>reproduce. Maybe some (more) other tb files have to be removed?
>>
>>I think that the same tablebases is not enough and
>>you may need to reprodce the all game with the same time of pondering
>>to reproduce the bug because
>>it is possible that learning from some previous analysis is responsible.
>>
>>The point of Amir is that not having tablebases is not the problem
>>and even if by some luck having all the tablebases could prevent the problem
>>then a similiar problem may happen even with all the tablebases.
>>>
>>>
>>>>The log doesn't show what J8 was thinking. Apparently it thought it was mating ,
>>>>because it played all moves in 0 seconds.
>>>
>>>
>>>Be5 is played here in several variations.
>>
>>
>>I dislike this move.
>>
>>This is a result of bad evaluation.
>>programs should translate mates from tablebases to numbers so in case of
>>a simple win they will not play stupid sacrifices.
>>
>>I have not statistics of the average number of moves to win based on
>>evaluation but a score of mate in 30 from tablebases should be transalted to
>>+8.00 in the evaluation if based on statistics of thousands of games without
>>tablebases
>>top programs always win positions with 7.50-8.50 score and the average
>>number of moves to win these positions games without tablebases is 30.
>>
>>Uri
>
>Some positions that are mate in 30 would yield a score of 1 or maybe even 0.5
>from some engines. Besides a mate in 30 is better than 7.5 or 9.


If 9 is practically always winning and usually in less than 30 moves than 9 is
better.


If the mate is
>trivial, the program should be able to solve it without the tbs. I'm sure there
>are some positions which the engine wouldn't be able to solve in time pressure
>meaning that lack of a vital tablebase would cost it half a point.

I talked about the ssdf games when there is no time pressure.

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.