Author: Uri Blass
Date: 14:47:34 04/01/02
Go up one level in this thread
On April 01, 2002 at 17:31:49, martin fierz wrote:
>On April 01, 2002 at 15:49:28, Roy Eassa wrote:
>
>>On April 01, 2002 at 14:36:35, Dieter Buerssner wrote:
>>
>>>On April 01, 2002 at 13:40:13, Roy Eassa wrote:
>>>
>>>>
>>>>[d] 8/8/5kpp/3K1p2/5P2/7P/6P1/8 b - - 0 1
>>>>Black to move and draw.
>>>>
>>>>I would think this would be trivially easy for today's programs running on fast
>>>>PCs, but (maybe) it isn't. There are only two first moves for Black that draw;
>>>>everything else loses:
>>>>
>>>>1...g5!
>>>>
>>>> [also drawing is 1...Kf7 2.Kd6
>>>> (2.h4 g5 3.h5 gxf4 4.Ke5 Ke7 5.Kxf5 Kf7 6.Kxf4 Kf6 7.g3 Ke6 =)
>>>> 2...Kf6 3.h4 g5 4.g3 Kg6 5.hxg5
>>>> (5.Ke6 g4=)
>>>> 5...hxg5 6.Ke6 gxf4 7.gxf4 Kg7 8.Kxf5 Kf7 =]
>>>>
>>>> [1...h5? 2.h4 +-
>>>> (2.g3 also wins)]
>>>>
>>>> [1...Ke7? 2.Ke5+-]
>>>>
>>>>2.g3 g4 3.h4
>>>>
>>>> [3.hxg4 fxg4 and Black pushes the pawn to h4 for a simple draw.]
>>>>
>>>>3...Kg6 4.Ke5 Kh5 =
>>>
>>>Yace needs a bit of time, to find g5.
>>>
>>> 1853091 36.172 -0.14 16. 1...h5 2. h4 Ke7 3. Ke5 Kf7 4. Kd6 Kf6 5. Kd7
>>> Kf7 6. g3 Kf6 7. Kd6 Kf7 8. Kc5 Ke7 9. g4 hxg4
>>> 10. Kc6 g3 11. Kb5 g2 12. h5
>>> 2262485 42.866 -0.54 17-- 1...h5 2. g3 Ke7 3. Ke5 Kf7 4. Kd6 Kf6 5. Kd7
>>> h4 6. gxh4 Kg7 7. Kd6 Kf7 8. Kc5 Kf6 9. Kd5 Kf7
>>> 10. Ke5 Ke8H 11. h5H g5H 12. fxg5H f4H {HT}
>>> 2543136 48.921 -0.55 17t 1...h5 2. g3 Ke7 3. Ke5 Kf7H 4. Kd6H Kf6H 5.
>>> Kd7H Kf7H 6. Kd8H Kf6H 7. Ke8H g5H 8. Kf8H h4H
>>> 9. gxh4H gxh4H 10. Kg8H Kg6H 11. Kh8H Kh5H 12.
>>> Kg8H Kh6H 13. Kf7H Kh7H {0}
>>> 3140092 1:09.2 -0.54 17t+ 1...g5 2. g3 g4 3. Kd6 Kg6 4. Ke5 Kh5 5. Kxf5
>>> gxh3 6. g4+ Kh4 7. Ke5 Kg3 8. f5 Kg2 9. f6 h2
>>> 10. f7
>>> 3435587 1:14.8 -0.24 17t 1...g5 2. g3 g4 3. h4 Kg6 4. Ke5 Kh5 5. Ke6H
>>> Kg6H 6. Ke7H Kh5H 7. Kd6H Kg6H 8. Kd5H Kh5H 9.
>>> Ke5H Kg6H 10. Kd4H Kh5H 11. Kc5H Kg6H 12. Kd5H
>>> Kh5H 13. Ke5H Kg6H 14. Kd4H Kh5H 15. Kc5H Kg6H
>>> 16. Kd5HR {0}
>>> 3797157 1:26.0 -0.24 17. 1...g5 2. g3 g4 3. h4 Kg6 4. Ke5 Kh5 5. Ke6 Kg6
>>> 6. Ke7 Kh5 7. Kd6 Kg6 8. Kd5 Kh5 9. Ke5 Kg6 10.
>>> Kd4 Kh5 11. Kc5 Kg6 12. Kd5 Kh5 13. Ke5 Kg6 14.
>>> Kd4 Kh5 15. Kc5 Kg6 16. Kd5 {0}
>>> 4408911 1:38.4 -0.24 18t 1...g5 2. g3 g4 3. h4 Kg6 4. Ke5 Kh5 5. Ke6 Kg6
>>> 6. Ke7 Kh5 7. Kd6 Kg6 8. Kd5 Kh5H 9. Ke6H Kg6H
>>> 10. Ke7H Kh5H 11. Kd6H Kg6H 12. Kd5H Kh5H 13.
>>>
>>>And essentially the same for some time. The fail low for h5 does not look
>>>impressive, but with the next depth, it would be much more negative. For Ke7,
>>>the behaviour is similar. It first looks reasonable, but soon fails low big
>>>time. If I disallow g5 (my "poor man's n-best mode) I get a similar score with
>>>Kf7 as the best move. All on K6-2 475.
>>>
>>>Regards,
>>>Dieter
>>
>>
>>
>>Is there ANY chess program (running on a fast PC) that can see, fairly quickly,
>>that 1...g5 draws AND that 1...h5 loses?
>>
>>(No offense, but showing -0.54 for the losing move and -0.24 for the drawing
>>move is not as decisive as one might imagine.)
>
>alpha-beta masks the true value yace might see here - it just sees that h5 is
>worse than g5, and does not care how much. you really need to have a different
>type of search (showing the n best moves with exact scores) to make this kind of
>comparison.
I believe that the score at depth 17 for h5 is an exact score.
It may see at depth 18 that h5 loses but the poster asked if there is a program
that can see quickly that 1...h5 loses and I guess that seeing it only at depth
18 after more than a minute(in case of 2 options) means not to see it quickly
Uri
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.