Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: fruit endgame evaluation

Author: Ryan B.

Date: 22:37:13 12/04/05

Go up one level in this thread


On December 05, 2005 at 01:31:21, allan johnson wrote:

>On December 04, 2005 at 20:15:42, Jaime Benito de Valle Ruiz wrote:
>
>>On December 04, 2005 at 19:08:08, allan johnson wrote:
>>
>>>Can someone tell me why Fruit2.2.1 evaluates this drawn position,
>>>[D]2b1k1b1/8/8/8/8/8/5K2/8 b - - 01
>>>so highly for black? (-7.99) Admittedly after the evaluation is completed an
>>>info message comes up saying insufficient material but I was curious about the
>>>reason for the score.
>>>Al
>>>
>>
>>The reason is very simple: Fruit uses a very simple code to recognize drawish
>>endgames, and in this case it assumes that a bishop pair will have both coloured
>>bishops (and I bet Fabien was aware of this limitation when he wrote the code).
>>He probably did it like this is because that is what you will get virtually
>>every single time in REAL games, so the code works fine in practice, not in
>>imaginary test positions like this.
>>Surely Fabien could have added a few more lines just to prevent people from
>>finding twisted positions that have no effect on the real playing strength of
>>the program, what why be bothered?
>>Regards,
>>
>>  Jaime
>Jaime Thanks for the feedback.You are probably right; it was just that I was
>surprised that Fritz 9 and CM 10 came up with the 0.00 score immediately.
>Al

Should not be surprised.  It is simple draw recog.  Not needed recog with EGTB
though.




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.