Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Draw detection(crafty) (more info)

Author: Jarkko Pesonen

Date: 12:29:00 02/04/03

Go up one level in this thread


The position was derived from this earlier position.
There is 9th move Nc3

[D] 7r/2qpkp2/p3p3/6P1/1p5r/7P/PPP2QK1/R2N1R2 b

Crafty 19.3

  7     0.12   1.93   2. ... Qc6+ 3. Qf3 Rxh3 4. Qxc6 dxc6
                      5. Nf2 Rh2+ 6. Kf3 Re8
  7->   0.25   1.93   2. ... Qc6+ 3. Qf3 Rxh3 4. Qxc6 dxc6
                      5. Nf2 Rh2+ 6. Kf3 Re8
  8     0.36   1.81   2. ... Qc6+ 3. Qf3 Rxh3 4. Qxc6 dxc6
                      5. Rf4 Rh2+ 6. Kf1 Rg8 7. Ne3 Rxg5
                      8. Rxb4
  8->   0.89   1.81   2. ... Qc6+ 3. Qf3 Rxh3 4. Qxc6 dxc6
                      5. Rf4 Rh2+ 6. Kf1 Rg8 7. Ne3 Rxg5
                      8. Rxb4
  9     1.04   1.47   2. ... Qc6+ 3. Qf3 Rxh3 4. Qxc6 dxc6
                      5. Nf2 R3h5 6. Kf3 Rxg5 7. Rad1
  9     5.67     ++   2. ... Rxh3!!
  9     8.11   0.79   2. ... Rxh3 3. Qxf7+ Kd8 4. Qf6+ Ke8
                      5. Kf2 Qxc2+ 6. Ke1 Qe4+ 7. Kd2 Qd3+
                      8. Kc1 Qc4+ 9. Nc3 bxc3
  9->   8.76   0.79   2. ... Rxh3 3. Qxf7+ Kd8 4. Qf6+ Ke8
                      5. Kf2 Qxc2+ 6. Ke1 Qe4+ 7. Kd2 Qd3+
                      8. Kc1 Qc4+ 9. Nc3 bxc3
 10    10.73   1.00   2. ... Rxh3 3. Qxf7+ Kd8 4. Qf6+ Ke8
                      5. Kf2 Qxc2+ 6. Ke1 Qe4+ 7. Kd2 Qd3+
                      8. Kc1 Qc4+ 9. Nc3 bxc3 10. g6 cxb2+
                      11. Kxb2
 10->  17.01   1.00   2. ... Rxh3 3. Qxf7+ Kd8 4. Qf6+ Ke8
                      5. Kf2 Qxc2+ 6. Ke1 Qe4+ 7. Kd2 Qd3+
                      8. Kc1 Qc4+ 9. Nc3 bxc3 10. g6 cxb2+
                      11. Kxb2
 11    23.64     ++   2. ... Rxh3!!
 11    34.36   0.56   2. ... Rxh3 3. Qxf7+ Kd8 4. Qf6+ Ke8
                      5. Kf2 Qxc2+ 6. Ke1 Qe4+ 7. Kd2 Qd3+
                      8. Kc1 Qc4+ 9. Nc3 bxc3 10. g6 cxb2+
                      11. Kxb2 Rh2+ 12. Ka3
 11->  42.58   0.56   2. ... Rxh3 3. Qxf7+ Kd8 4. Qf6+ Ke8
                      5. Kf2 Qxc2+ 6. Ke1 Qe4+ 7. Kd2 Qd3+
                      8. Kc1 Qc4+ 9. Nc3 bxc3 10. g6 cxb2+
                      11. Kxb2 Rh2+ 12. Ka3
 12     1:18     ++   2. ... Rxh3!!
 12     3:11   0.00   2. ... Rxh3 3. Qxf7+ Kd8 4. Qf6+ Ke8
                      5. Kf2 Qxc2+ 6. Ke1 Qe4+ 7. Kd2 Qd3+
                      8. Ke1 Re3+ 9. Nxe3 Qxe3+ 10. Kd1 Qd3+
                      11. Ke1 Qe3+


Ruffian:
  Depth   Time  Score    Best line
   8/23   1.51     ++    ... f5
   8/23   2.53  -1.21    ... f5 Rg1 Rxh3 Qd4 Qxc2+ Kf1 Ke8 Ke1 a5
   8-->   2.73  -1.21    ... f5 Rg1 Rxh3 Qd4 Qxc2+ Kf1 Ke8 Ke1 a5
   9/24   3.76  -1.31    ... f5 Rg1 Rxh3 Kf1 Rh2 Qd4 Qxc2 Qxb4+ Ke8 Qb8+ Ke7
                         Qe5
   9/25   5.42   x.xx    ... Rxh3
   9/25   5.95     ++    ... Rxh3
   9/26   7.06   0.00    ... Rxh3 Qxf7+ Kd8 Qf6+ Ke8 Kf2 Qxc2+ Ke1 Qe4+ Kd2
                         Qd3+ Kc1 Qc4+ Kb1 Qe4+ Kc1 Qc4+
   9-->   7.53   0.00    ... Rxh3 Qxf7+ Kd8 Qf6+ Ke8 Kf2 Qxc2+ Ke1 Qe4+ Kd2
                         Qd3+ Kc1 Qc4+ Kb1 Qe4+ Kc1 Qc4+
  10/28   8.92   0.00    ... Rxh3 Qxf7+ Kd8 Qf6+ Ke8 Kf2 Qxc2+ Ke1 Qe4+ Kd2
                         Qd3+ Kc1 Qc4+ Kb1 Qe4+ Kc1 Qc4+
  10-->   9.87   0.00    ... Rxh3 Qxf7+ Kd8 Qf6+ Ke8 Kf2 Qxc2+ Ke1 Qe4+ Kd2
                         Qd3+ Kc1 Qc4+ Kb1 Qe4+ Kc1 Qc4+
  11/31  14.73   0.00    ... Rxh3 Qxf7+ Kd8 Qf6+ Ke8 Kf2 Qxc2+ Ke1 Qe4+ Kd2
                         Qd3+ Kc1 Qc4+ Kb1 Qe4+ Kc1 Qc4+
  11-->  17.62   0.00    ... Rxh3 Qxf7+ Kd8 Qf6+ Ke8 Kf2 Qxc2+ Ke1 Qe4+ Kd2
                         Qd3+ Kc1 Qc4+ Kb1 Qe4+ Kc1 Qc4+


There is difference in speed an in which ply depth
draw is detected

Jarkko

******************************************

On February 04, 2003 at 13:28:55, Robert Hyatt wrote:

>On February 04, 2003 at 13:26:43, Robert Hyatt wrote:
>
>>On February 04, 2003 at 08:37:57, Jarkko Pesonen wrote:
>>
>>>[D] 4k2r/3p4/p3pQ2/6P1/1pq5/7r/PP6/R1KN1R2 w
>>>
>>>This is draw , but how do you statically can evaluate that it so?
>>>
>>>This was the quiescent position and crafty suggested Nc3 bxc3 and not go for
>>>repetion(eval 0.09 for crafty statically after these moves but search from this
>>>position reveals that this loses rook at lest ) and ruffian correctly wanted to
>>>play Kb1 which was the repeating move.
>>>
>>>
>>>Anyway ruffian seems to find drawing lines much sooner than crafty.
>>>Does anyone have a glue why it is so fast.
>>>It catches draw a few ply earlier than crafty.
>>>Is there a way fix this in crafty?
>>>
>>>
>>>Jarkko
>>
>>
>>You are probably mis-interpreting the concept of "draw".
>>
>>For a program to see a draw by repetition, two things have to happen:
>>
>>1.  It has to believe that one side can _force_ the repetition, if the other
>>side plays
>>in a particular way.
>>
>>2.  It has to believe that the other side would be worse off it it tries to
>>avoid the
>>repetition.
>>
>>All you need in a position like this is to think that one side is is better than
>>the other,
>>and you might or might not be able to see the repetition.  IF one program thinks
>>white is slightly better, then it would think that black would try to force the
>>repetition
>>and white must go along or else get "worse".  If another program thinks black is
>>better, then it might totally ignore the repetition opportunity.
>>
>>Usually one side has the option to vary, say in a perpetual check the checking
>>side doesn't
>>have to check.  But he has to believe that if he doesn't check, the opponent has
>>a better
>>resulting position.  If that is not true, the perpetual won't be played.
>
>
>I tried this position and Crafty spots a repetition instantly:
>
>                5     0.04 -Mat04   1. Kd2 Rd3+ 2. Ke1 Qe4+ 3. Kf2 Rh2+
>                                    4. Kg1 Qh1#
>                5     0.07  -0.46   1. Nc3 bxc3 2. g6 cxb2+ 3. Kxb2 d6
>                5     0.07   0.00   1. Kb1 Qd3+ 2. Kc1 Qc4+
>                7     0.15   0.00   1. Kb1 Qd3+ 2. Kc1 Qc4+
>                7->   0.16   0.00   1. Kb1 Qd3+ 2. Kc1 Qc4+
>                8     0.33   0.00   1. Kb1 Qd3+ 2. Kc1 Qc4+
>                8->   0.42   0.00   1. Kb1 Qd3+ 2. Kc1 Qc4+ (s=2)
>                9     0.93   0.00   1. Kb1 Qd3+ 2. Kc1 Qc4+
>                9->   1.01   0.00   1. Kb1 Qd3+ 2. Kc1 Qc4+
>               10     2.34   0.00   1. Kb1 Qd3+ 2. Kc1 Qc4+
>               10->   2.38   0.00   1. Kb1 Qd3+ 2. Kc1 Qc4+
>               11     7.25   0.00   1. Kb1 Qd3+ 2. Kc1 Qc4+
>               11->   7.68   0.00   1. Kb1 Qd3+ 2. Kc1 Qc4+
>               12    29.98   0.00   1. Kb1 Qd3+ 2. Kc1 Qc4+
>               12->  30.48   0.00   1. Kb1 Qd3+ 2. Kc1 Qc4+
>
>
>I'm not sure it could do any better than .07 seconds for spotting this, and this
>was not
>run on my dual xeon either, but on a quad 550 that is about 1/3 the speed...



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.