Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: a bug in Fritz5 and Junior4.6

Author: Bernhard Bauer

Date: 23:45:39 04/07/98

Go up one level in this thread


On April 08, 1998 at 01:56:58, Christophe Theron wrote:

>On April 06, 1998 at 09:49:25, blass uri wrote:
>
>>
>>On April 06, 1998 at 05:33:45, Sylvain Renard wrote:
>>
>>>
>>>On April 06, 1998 at 03:35:06, Roland Pfister wrote:
>>>>Patzer found Qe3 in 3:08 but it thinks it is a mate in 15 :-(
>>>>I thought I had found that bug...
>>>  Capture plays Qe3 very quickly and see mate in 9 in less than
>>>2 minutes (P PRO 200 MHz). The bug of Fritz 5 is vey strange,
>>>someone as an idea of its origine?
>>>   Sylvain
>>
>>The evaluation function of fritz5 was:
>>6.09 9/18
>> 5.78 10/20
>> 5.47 11/22
>> 5.16 12/25
>> 4.84 13/25
>>4.53 14/28
>>4.22 15/30
>>3.91 16/32
>>3.59 17/34
>>3.28 18/36
>>2.97 19/38
>>I think this is something like 9-0.31*d when d is the depth of bruth
>>force
>>maybe fritz5 do some average between its evaluation by brute force
>>and the eveluation of not brute force and use the depth.
>>fritz4 and fritz3 have similiar bug fritz3 go almost to 0 but then go
>>back
>>in depth 16 it's evaluation is 0.13 and in depth 17 is 0.44 (in depth 15
>>0.44
>>and in depth 14 0.75)
>
>I think it's a zugzwang problem. I don't know exactly why, but it looks
>like the black queen has to stay on the a1-h8 diagonal (or else the
>white queen goes on this diagonal, check, and mate in 2), and has also
>to stay on the the a file.
>
>Or something like that. Anyway, I think black is quickly caught in
>zugzwang.
>
>Programs relying on null move like Fritz (and maybe Crafty?) have
>problems with this, because a general rule in those programs is that
>they don't care of zugzwang problem if the queens are on the board. This
>is correct 99.99% of the time, but not in this position...
>

Here is some output from crafty:

White(1): hash table memory = 6M bytes.
White(1): pawn hash table memory = 1.3M bytes.

       +---+---+---+---+---+---+---+---+
    8  |   |   |   |   |   |   |   | *K|
       +---+---+---+---+---+---+---+---+
    7  |   |   |   |   | K |   |   | *P|
       +---+---+---+---+---+---+---+---+
    6  |   |   |   |   |   |   |   | P |
       +---+---+---+---+---+---+---+---+
    5  |   |   |   | *P|   |   | Q |   |
       +---+---+---+---+---+---+---+---+
    4  |   |   |   |   |   |   |   |   |
       +---+---+---+---+---+---+---+---+
    3  |   |   |   |   |   |   |   |   |
       +---+---+---+---+---+---+---+---+
    2  |   |   |   |   |   |   |   |   |
       +---+---+---+---+---+---+---+---+
    1  | *Q|   |   |   |   |   |   |   |
       +---+---+---+---+---+---+---+---+
         a   b   c   d   e   f   g   h

White(1): end-game phase
              clearing hash tables
              time surplus   0.00  time limit 5:00 (5:00)
              depth   time  score   variation (1)
                6->   0.84   0.29   1. Qf4 Qa7+ 2. Ke6 Qb6+ 3. Kxd5 Qb3+
                                    4. Kc5 Qe6 5. Qf8+
                7->   1.79   0.00   1. Qf4 Qa7+ 2. Ke6 Qb6+ 3. Kxd5 Qb5+
                                    4. Ke6 Qe8+ 5. Kf6 Qf8+ 6. Kg5 Qxf4+
                                    7. Kxf4
                8->   3.70   0.12   1. Qe3 Qb2 2. Ke6 Qa1 3. Qb6 Qe1+ 4.
                                    Kxd5 Qd2+ 5. Ke5 Qd3
                9->   5.00   0.22   1. Qe3 Qb2 2. Ke6 Qa1 3. Qf4 Qa6+ 4.
                                    Ke5 Kg8 5. Kxd5 Qb6
               10->   9.78   0.00   1. Qe3 Qb2 2. Ke6 Qa1 3. Qf4 Qa6+ 4.
                                    Kxd5 Qb5+ 5. Ke6 Qe8+ 6. Kf6 Qf8+ 7.
                                    Kg5 Qxf4+ 8. Kxf4
               11->  16.43   0.00   1. Qe3 Qb2 2. Ke6 Qa1 3. Qf4 Qa6+ 4.
                                    Kxd5 Qb5+ 5. Ke6 Qe8+ 6. Kd5 Qf7+ 7.
                                    Qxf7
               12    22.23     ++   1. Qe3!!
               12    29.48  Mat09   1. Qe3 Qb2 2. Ke6 Qa1 3. Qb6 Qg7 4.
                                    hxg7+ Kxg7 5. Kxd5 <EGTB>
               12->  47.53  Mat09   1. Qe3 Qb2 2. Ke6 Qa1 3. Qb6 Qg7 4.
                                    hxg7+ Kxg7 5. Kxd5 <EGTB>
              time: 47.54  cpu:0%  mat:-1  n:4443263  nps:93444
              ext-> checks:825089 recaps:8581 pawns:15123 1rep:134501
              predicted:0  nodes:4443263  evals:370831
              endgame tablebase-> probes done: 37303  successful: 37303
              hashing-> trans/ref:63%  pawn:99%  used:w93% b99%

Crafty will mate in 9 moves.

So crafty seems to be pretty good on this problem.
Kind regards
B. Bauer

White(1): Qe3
              time used:  47.56
Black(1): execution complete.
>More help wanted... Can somebody prove that black is in zugzwang in some
>important variant of this position?
>
>
>    Christophe



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.