Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Importance of TBs once more!(an example to mistake of computer)

Author: blass uri

Date: 08:23:41 10/13/98

Go up one level in this thread



On October 13, 1998 at 07:34:46, Bernhard Bauer wrote:

>On October 13, 1998 at 06:05:55, blass uri wrote:
>
>>
>>On October 13, 1998 at 05:54:00, blass uri wrote:
>>
>>>
>>>On October 13, 1998 at 04:34:00, Amir Ban wrote:
>>>
>>>>On October 13, 1998 at 02:12:24, Jouni Uski wrote:
>>>>
>>>>>In SSDF game archive there is 5834 games. In 21 games there is KQ-KR ending.
>>>>>And 20 of these were won by Q side WITHOUT any use of TBs! Quite surprising -
>>>>>isn't this normally considered as difficult ending?
>>>>
>>>>I've never seen this ending in a computer-computer game where the side with the
>>>>queen failed to win, at any time control.
>>>>
>>>>This endgame is quite easy for a computer.
>>>
>>>Is it easy when the defendor use tablebases?
>>>
>>>I found that the defendor without tablebases(when the defendor is a computer
>>>program) may miss a simple draw by perpetual check in positions when there is
>>>black queen at f2 and white king at h1.
>>
>>Here is an example
>>
>>7R/8/8/8/8/k7/5q2/7K white to move and draw
>>Fritz5(without tablebases) plays Rg8 and Junior5(16 bit) Rh3+
>>when Ra8+ is the only way to draw.
>>
>>Uri
>>
>
>Crafty 15.20 finds the draw without tablebases.
>On a 2xPPro 233MHz crafty sees Ra8+ at ply 14 in 2:48 min.
>Crafty sees the draw (score =-0.66) at ply 23 in 3:45 min.
>The output for ply 42 is:
>               42->   8:04  -0.66   1. Ra8+ Kb3 2. Rb8+ Kc3 3. Rc8+ Kd3
>                                    4. Rd8+ Ke3 5. Re8+ Kd4 6. Rd8+ Ke4
>                                    7. Re8+ Kd5 8. Rd8+ Kc4 9. Rc8+ Kb4
>                                    10. Rb8+ Ka4 11. Ra8+ Kb5 12. Rb8+
>                                    Ka5 13. Ra8+ Kb6 14. Rb8+ Kc5 15. Rc8+
>                                    Kd6 16. Rd8+ Kc6 17. Rc8+ Kd7 18. Rd8+
>                                    Ke6 19. Re8+ Kd5
>Kind regards
>Bernhard


You need to search at least 54 plies forward to see the stalemate or repetition
and many programs cannot search more than 50 plies.

I think that most of the top programs cannot search to 42 plies in this position
even in 8000 minutes.
I think that crafty is more than 1000 times faster than most of the top programs
in finding the draw.

Uri

>
>
>>
>>>
>>>Uri
>>>>
>>>>Amir



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.