Author: Omid David Tabibi
Date: 13:16:55 08/21/03
Go up one level in this thread
On August 21, 2003 at 16:09:42, Dieter Buerssner wrote:
>On August 21, 2003 at 16:02:35, Omid David Tabibi wrote:
>
>>I did say that Junior probably gains considerably by this assumption. My
>>objection is a matter of principle: given enough time a program should play
>>correctly in any (practical) position.
>
>Exactly the same here. Omid, when I read your paper about null move pruning, I
>thought, it could fail for this objection. Did I read it wrong.
>
>A position discussed in the German CSS-Forum:
>
>[D] 8/1B6/8/5p2/8/8/5Qrq/1K1R2bk w - -
4 seconds on my 733 MHz system:
depth time nodes score variation
1/ 7-> 0.02 0k 2.15 1.b7g2 h2g2 2.f2f5
2/ 9-> 0.02 0k 2.22 1.d1f1 f5f4 2.b7g2 h2g2 3.f2g2 h1g2
4.f1f4
3/11-> 0.02 1k 3.10 1.f2f5 h2g3 2.f5h5
4/11-> 0.03 3k 3.10 1.f2f5 h2g3 2.f5h5 g3h2 3.b7g2 h1g2
5/13-> 0.06 12k 2.24 1.f2f3 f5f4 2.d1d2 g1f2 3.f3f2
6/13 0.07 14k 2.30 1.f2f3 f5f4 2.d1d2 g1f2 3.d2f2 h2g1
4.b1a2
6/14 0.09 19k ++ 1.f2f5
6/15 0.15 38k 7.02 1.f2f5 h2h6 2.b7g2 h1h2 3.d1d6 h6e3
4.g2f3
6/15-> 0.17 41k 7.02 1.f2f5 h2h6 2.b7g2 h1h2 3.d1d6 h6e3
4.g2f3
7/16 0.25 75k -- 1.b1c2
7/19 0.49 136k -6.53 1.b1c2 h2c7 2.c2d3 c7b7 3.d1g1 g2g1
4.f2h4 h1g2 5.h4g5 g2f2 6.g5f5
7/19 0.70 193k 6.52 1.f2f5 h2h6 2.b7g2 h1h2 3.d1d2 g1d4
4.f5d3 h6e3
7/19-> 0.71 195k 6.52 1.f2f5 h2h6 2.b7g2 h1h2 3.d1d2 g1d4
4.f5d3 h6e3
8/19 0.88 247k ++ 1.f2f5
8/19 1.19 362k 7.13 1.f2f5 h2h6 2.b7g2 h1h2 3.g2c6 h6c6
4.f5h5 h2g2 5.h5g4 g2h2 6.d1g1
8/19-> 1.46 456k 7.13 1.f2f5 h2h6 2.b7g2 h1h2 3.g2c6 h6c6
4.f5h5 h2g2 5.h5g4 g2h2 6.d1g1
9/21 2.24 716k 7.31 1.f2f5 h2g3 2.f5h7 g3h2 3.b7g2 h1g2
4.d1d2 g2g3 5.h7d3 g3g4 6.d2h2 g1h2
7.b1c2
9/21 2.45 801k ++ 1.f2f3
9/21 2.62 858k 9.22 1.f2f3 f5f4 2.d1d2 g1f2 3.d2f2 h2g3
4.f3h5 h1g1 5.f2g2 g3g2 6.b7g2 g1g2
9/21 4.40 1643k 18.26 1.f2d4 f5f4 2.b1a1 f4f3 3.b7f3 h2e5
4.d1g1 h1h2 5.g1g2 h2h3 6.d4e5
9/21 4.65 1746k #3 1.f2a7 f5f4 2.b7f3 h2h5 3.a7g1
>
>A mate in 3. My engine, using a different sort of null move pruning (that
>theoretically really will play correctly in any position given enough time),
>needed over 10 minutes ...
>
>An engine can be lucky (or clever) with the evals, and null move pruning may not
>avoid finding the mate fast in the above position. In general, however the above
>position will be unsolvable for normal null move algorithms.
>
>Regards,
>Dieter
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.