Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: How long to see black is winning + crafty bench

Author: Mike Byrne

Date: 07:24:52 03/27/05

Go up one level in this thread


On March 27, 2005 at 06:56:35, Vincent Lejeune wrote:

>On March 26, 2005 at 21:30:04, Robert Hyatt wrote:
>
>>On March 26, 2005 at 19:46:58, Vincent Lejeune wrote:
>>
>>>On March 26, 2005 at 15:34:45, Dieter Buerssner wrote:
>>>
>>>>On March 26, 2005 at 15:29:42, Dieter Buerssner wrote:
>>>>
>>>>>On March 26, 2005 at 15:20:17, Mike Byrne wrote:
>>>>>
>>>>>>after the main line :  Kf2 f5 h3 Kc7 g3 hxg3 is winning no?
>>>>>>
>>>>>>Black(2): his
>>>>>>    white       black
>>>>>>  1  Kf2         f5
>>>>>>  2  h3          Kc7
>>>>>>  3  g3          ...
>>>>>>               17    25.52  -7.17   3. ... hxg3+ 4. Kxg3 f4+ 5. Kg4 b6
>>>>>>                                    6. h4 f3 7. Kxf3 gxh4 8. c6 Kxc6 9.
>>>>>>                                    Kg4 b5 10. a5 b4 11. Kxh4 Kb5 12. Kg4
>>>>>>
>>>>>
>>>>>In this line, Yace says 8. c6? is losing, while 8.cxb6 is draw.
>>>>
>>>>It is such an obvious blunder, that it looks like a bug in Crafty.
>>>>
>>>>Regards,
>>>>Dieter
>>>
>>>
>>>Probably a bug in the static eval of the pawn endgame. after 3 min@3 Ghz :
>>
>>I can't reproduce Mikes PV with this "c6 bug"...  I don't get c6 at any point
>>that I can see, it always takes...
>>
>>Any ideas?
>
>
>I succeded to replicate c6 as best move :
>1) without any EGTB
>2) starting with this position 8/1p1k4/5p2/2Pp2p1/P2P3p/8/6PP/6K1 w - - 0 1
>3) under chessbase interface
>4) played 1. Kf2 f5 2. h3 Kc7 3. g3 hxg3+ 4. Kxg3 with 5 between each move
>5) the score is rising for white but it so long I stopped
>Is it a problem related to only pawns endgame eval or linked to hashtables too ?
>

I suggest not to test under Chessbase GUI unless you are using the native
chessbase crafty engine. Chessbasee requires an UCI engine, Crafty is not a UCI
engine.  If the "bug" is only in the Chessbase interface and not in the console
window, I beleives  that may suggest that there in bug Chessbase GUI
implementation.   Adding an adapater in between to make Crafty works adds
another level of debugging complexitiy.

In my case, I have isolated the issue to the kppkp EGTB that I generated.  They
were generated using the vector option, I am now generating them without the
vector option to see what happens.  One obvious difference is that the vector
and non-vector generated kppkp files have different file sizes which may be
indicator of a bug in my compiled version of tbgen.

Best,

Michael





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.