Author: Robert Hyatt
Date: 11:28:49 04/20/04
Go up one level in this thread
On April 20, 2004 at 06:10:05, Omid David Tabibi wrote:
>On April 19, 2004 at 22:50:56, Robert Hyatt wrote:
>
>>On April 19, 2004 at 15:49:14, Andrew Williams wrote:
>>
>>>On April 18, 2004 at 19:56:25, Andrew Wagner wrote:
>>>
>>>Try searching longer. PM is getting this at ply 27 (78kns, 0.76 seconds) at the
>>>moment, although I've seen it return the correct move at various different plies
>>>over the years.
>>>
>>>Andrew
>>
>>Something might be wrong. It is a 26 ply combination to win a pawn, with no
>>hash table at all... Hashing usually lets us hit the solution quicker. IE
>>Crafty typically solves it at depth=18...
>
>In his article "PEASANT: An endgame program for kings and pawns", Newborn
>writes: "Position 70 would require a 30-ply search (25,000 hours)"
I verified that in the "Chess Skill in Man and Machine." Unfortunately, someone
ran off with my "Basic Chess Endings" book many years ago so I can't find out
what Fine offered as analysis, but Monty is probably quoting him and his book
(Fine's) had more than enough errors to let me believe he missed this one a bit.
He got the right first move, and the reason it works, but he probably missed a
deep zugzwang that ends this on ply 27 rather than 31...
Here is some Crafty analysis. Hashing is enabled but no scores/bounds are used.
I still use the hash move for ordering, however, but that is all. That makes
this go slow. I ran this on my dual 2.8 using both CPUs...
To be compatible with Dieter, I also used a material-only evaluation, and Crafty
finds Kb1 at depth=26, which seems to confirm that this is the right depth.
Only took five minutes which shows how fast today's hardware really is. :) I
had no idea this could be solved without hashing, that quickly.
25 54.81 1.00 1. Kb2 Ka6 2. Kc3 Kb7 3. Kc4 Kb6 4.
Kd3 Kc7 5. Ke2 Kd7 6. Kf3 Ke7 7. Kg3
Kf6 8. Kh4 Kg6 9. Kh3 Kh5 10. Kg3 Kg6
11. Kf3 Kh5 12. Kf2 Kg6 13. Ke3 (s=2)
25-> 1:07 1.00 1. Kb2 Ka6 2. Kc3 Kb7 3. Kc4 Kb6 4.
Kd3 Kc7 5. Ke2 Kd7 6. Kf3 Ke7 7. Kg3
Kf6 8. Kh4 Kg6 9. Kh3 Kh5 10. Kg3 Kg6
11. Kf3 Kh5 12. Kf2 Kg6 13. Ke3
26 3:24 1.00 1. Kb2 Ka8 2. Kc3 Kb7 3. Kb3 Kc7 4.
Kc4 Kb6 5. Kd3 Kc7 6. Ke3 Kd7 7. Kf3
Ke7 8. Kg3 Kf6 9. Kh4 Kg6 10. Kh3 Kh5
11. Kg3 Kg6 12. Kf3 Kf6 13. Kg3 Kg6
26 5:09 +1 1. Kb1!!
26 5:56 2.00 1. Kb1 Ka8 2. Kb2 Ka7 3. Kb3 Ka6 4.
Kc2 Kb7 5. Kc3 Kb6 6. Kc4 Ka6 7. Kd3
Kb7 8. Ke3 Kb6 9. Kf3 Ka6 10. Kg3 Kb7
11. Kh4 Kb6 12. Kg5 Ka6 13. Kxf5 Ka7
26-> 5:56 2.00 1. Kb1 Ka8 2. Kb2 Ka7 3. Kb3 Ka6 4.
Kc2 Kb7 5. Kc3 Kb6 6. Kc4 Ka6 7. Kd3
Kb7 8. Ke3 Kb6 9. Kf3 Ka6 10. Kg3 Kb7
11. Kh4 Kb6 12. Kg5 Ka6 13. Kxf5 Ka7
(s=3)
Same machine, normal crafty:
23-> 0.00 1.21 1. Kb1 Ka8 2. Kb2 Kb7 3. Kc3 Kc7 4.
Kd3 Kb7 5. Kc4 Kb6 6. Kb3 Kc7 7. Kc2
Kc8 8. Kd2 <HT>
24 0.00 +1 1. Kb1!!
24 0.00 +3 1. Kb1!!
24 0.01 3.54 1. Kb1 Kb7 2. Kc1 Kb8 3. Kc2 Kc8 4.
Kd2 Kc7 5. Kd3 Kb7 6. Ke3 Kc7 7. Kf3
Kd7 8. Kg3 Ke7 9. Kh4 Kf6 10. Kh5 Kf7
11. Kg5 Kg7 12. Kxf5 Kf7
24-> 0.01 3.54 1. Kb1 Kb7 2. Kc1 Kb8 3. Kc2 Kc8 4.
Kd2 Kc7 5. Kd3 Kb7 6. Ke3 Kc7 7. Kf3
Kd7 8. Kg3 Ke7 9. Kh4 Kf6 10. Kh5 Kf7
11. Kg5 Kg7 12. Kxf5 Kf7
25 0.02 3.60 1. Kb1 Kb7 2. Kc1 Kb8 3. Kc2 Kc8 4.
Kd2 Kc7 5. Kd3 Kb7 6. Ke3 Kc7 7. Kf3
Kd7 8. Kg3 Ke7 9. Kh4 Kf6 10. Kh5 Kf7
11. Kg5 Ke7 12. Kxf5 Kd7 13. Ke4
0.00 seconds to find the winning move. .03 seconds to finish depth=26...
hashing makes a huge difference.
>
>But running the position on Falcon:
>
>Falcon 1.0.12.6 running on GenuineIntel 733MHz 256MB:
>depth time nodes nps score variation
>18/18 0.12 17k 149k 0.76 1.a1b2 a7a6 2.b2c3 a6b7 3.c3d3 b7c7
> 4.d3e3 c7d7 5.e3d2 d7c7 6.d2d3 c7b6
> 7.d3c2 b6b7 8.c2c3 b7c7 9.c3c4 c7d7
>19/20 1.05 284k 270k 0.76 1.a1b2 a7a8 2.b2c3 a8b7 3.c3d3
>19/20 1.05 284k 270k 0.86 1.a1b1++
>19/20 1.06 284k 268k 1.92 1.a1b1++
>19/20 2.58 744k 288k 2.36 1.a1b1 a7b7 2.b1c1 b7c7 3.c1d1 c7d7
> 4.d1c2 d7c8 5.c2d2 c8d7 6.d2c3 d7e7
> 7.c3c4 e7d7 8.c4b5 d7c7 9.b5a5 c7b7
> 10.a5b4
>
>It takes a mere 1 second at depth 19!
>
>
>
>>
>>
>>>
>>>> Ok, I've been banging my head against this one all day. This is Fine #70:
>>>>[d]8/k7/3p4/p2P1p2/P2P1P2/8/8/K7 w - -
>>>>
>>>> Below is Trueno's output for the position. Of course, the correct result is
>>>>+1.8 (Trueno counts pawns as .9) with the best move of Kb1. It finds that but
>>>>then changes its mind. I've been told that this position is generally infamous
>>>>for catching hash bugs. Anyone have any suggestions how to attack this one? Null
>>>>move is off, and I'm evaluating material only, using a single replace-always
>>>>hash table. Any suggestions would be VERY much appreciated). Andrew
>>>>
>>>> -------------------------------
>>>> 8 | . | . | . | . | . | . | . | . |
>>>> |-------------------------------|
>>>> 7 | k | . | . | . | . | . | . | . |
>>>> |-------------------------------|
>>>> 6 | . | . | . | p | . | . | . | . |
>>>> |-------------------------------|
>>>> 5 | p | . | . | P | . | p | . | . |
>>>> |-------------------------------|
>>>> 4 | P | . | . | P | . | P | . | . |
>>>> |-------------------------------|
>>>> 3 | . | . | . | . | . | . | . | . |
>>>> |-------------------------------|
>>>> 2 | . | . | . | . | . | . | . | . |
>>>> |-------------------------------|
>>>> 1 | K | . | . | . | . | . | . | . |
>>>> -------------------------------
>>>> A B C D E F G H
>>>>
>>>>Fifty: 0 En Passant: 0
>>>>Castle: 0000
>>>>Ply: 0 To move: True
>>>>
>>>>
>>>>1. Score: .9; Time: 0; nodes: 4; PV: a1-b2
>>>>2. Score: .9; Time: 0; nodes: 15; PV: a1-b2 a7-b6
>>>>3. Score: .9; Time: 0; nodes: 48; PV: a1-b2 a7-b6 b2-c3
>>>>4. Score: .9; Time: 0; nodes: 111; PV: a1-b2 a7-b6 b2-c3 b6-c7
>>>>5. Score: .9; Time: 0; nodes: 231; PV: a1-b2 a7-b6 b2-c3 b6-c7 c3-c4
>>>>6. Score: .9; Time: 0; nodes: 408; PV: a1-b2 a7-b6 b2-c3 b6-c7 c3-c4 c7-d7
>>>>7. Score: .9; Time: 0; nodes: 671; PV: a1-b2 a7-b6 b2-c3 b6-c7 c3-c4 c7-d7 c4-d3
>>>>8. Score: .9; Time: 0; nodes: 1003; PV: a1-b2 a7-b6 b2-c3 b6-c7 c3-c4 c7-d7
>>>>c4-d3 d7-e7
>>>>9. Score: .9; Time: 0; nodes: 1443; PV: a1-b2 a7-b6 b2-c3 b6-c7 c3-c4 c7-d7
>>>>c4-d3 d7-e7 d3-c4
>>>>10. Score: .9; Time: 0; nodes: 1968; PV: a1-b2 a7-b6 b2-c3 b6-c7 c3-c4 c7-d7
>>>>c4-d3 d7-e7 d3-c4 e7-d7
>>>>11. Score: .9; Time: 0; nodes: 2735; PV: a1-b2 a7-b6 b2-c3 b6-c7 c3-c4 c7-d7
>>>>c4-d3 d7-e7 d3-c4 e7-d7 c4-d3
>>>>12. Score: .9; Time: 0; nodes: 3522; PV: a1-b2 a7-b6 b2-c3 b6-c7 c3-c4 c7-d7
>>>>c4-d3 d7-e7 d3-c4 e7-d7 c4-d3 d7-e7
>>>>13. Score: .9; Time: 0; nodes: 4416; PV: a1-b2 a7-b6 b2-c3 b6-c7 c3-c4 c7-d7
>>>>c4-d3 d7-e7 d3-c4 e7-d7 c4-d3 d7-e7 d3-c4
>>>>14. Score: .9; Time: 1; nodes: 5410; PV: a1-b2 a7-b6 b2-c3 b6-c7 c3-c4 c7-d7
>>>>c4-d3 d7-e7 d3-c4 e7-d7 c4-d3 d7-e7 d3-c4 e7-d7
>>>>15. Score: .9; Time: 1; nodes: 6584; PV: a1-b2 a7-b6 b2-c3 b6-c7 c3-c4 c7-d7
>>>>c4-d3 d7-e7 d3-c4 e7-d7 c4-d3 d7-e7 d3-c4 e7-d7 c4-d3
>>>>16. Score: .9; Time: 1; nodes: 7885; PV: a1-b2 a7-b6 b2-c3 b6-c7 c3-c4 c7-d7
>>>>c4-d3 d7-e7 d3-c4 e7-d7 c4-d3 d7-e7 d3-c4 e7-d7 c4-d3 d7-e7
>>>>17. Score: .9; Time: 1; nodes: 9273; PV: a1-b2 a7-b6 b2-c3 b6-c7 c3-c4 c7-d7
>>>>c4-d3 d7-e7 d3-c4 e7-d7 c4-d3 d7-e7 d3-c4 e7-d7 c4-d3 d7-e7 d3-c4
>>>>18. Score: .9; Time: 1; nodes: 10845; PV: a1-b2 a7-b6 b2-c3 b6-c7 c3-c4 c7-d7
>>>>c4-d3 d7-e7 d3-c4 e7-d7 c4-d3 d7-e7 d3-c4 e7-d7 c4-d3 d7-e7 d3-c4 e7-d7
>>>>19. Score: .9; Time: 1; nodes: 16249; PV: a1-b2 a7-b6 b2-c3 b6-b7 c3-c4 b7-b6
>>>>c4-d3 b6-c7 d3-c4 c7-d7 c4-d3 d7-e7 d3-c4 e7-d7 c4-d3 d7-e7 d3-c4 e7-d7 c4-d3
>>>>20. Score: .9; Time: 1; nodes: 18816; PV: a1-b2 a7-b6 b2-c3 b6-b7 c3-c4 b7-b6
>>>>c4-d3 b6-c7 d3-c4 c7-d7 c4-d3 d7-e7 d3-c4 e7-d7 c4-d3 d7-e7 d3-c4 e7-d7 c4-d3
>>>>d7-e7
>>>>21. Score: 1.8; Time: 2; nodes: 30305; PV: a1-b1 a7-b6 b1-c2 b6-b7 c2-d2 b7-c7
>>>>d2-d3 c7-b6 d3-e3 b6-c7 e3-f3 c7-d7 f3-g3 d7-e7 g3-h4 e7-f6 h4-h5 f6-f7 h5-g5
>>>>f7-e8 g5-f5
>>>>22. Score: 1.8; Time: 2; nodes: 40127; PV: a1-b1 a7-b6 b1-c2 b6-b7 c2-d2 b7-c7
>>>>d2-d3 c7-b6 d3-e3 b6-c7 e3-f3 c7-d7 f3-g3 d7-e7 g3-h4 e7-f6 h4-h5 f6-f7 h5-g5
>>>>f7-e8 g5-f5 e8-e7
>>>>23. Score: 1.8; Time: 2; nodes: 50869; PV: a1-b1 a7-b6 b1-c2 b6-b7 c2-d2 b7-c7
>>>>d2-d3 c7-b6 d3-e3 b6-c7 e3-f3 c7-d7 f3-g3 d7-e7 g3-h4 e7-f6 h4-h5 f6-f7 h5-g5
>>>>f7-e8 g5-f5 e8-e7 f5-e4
>>>>24. Score: 1.8; Time: 3; nodes: 64039; PV: a1-a2 a7-b6 a2-b3 b6-c7 b3-c3 c7-b6
>>>>c3-c2 b6-a6 c2-d2 a6-b6 d2-e3 b6-c7 e3-f3 c7-d7 f3-g3 d7-e7 g3-h4 e7-f6 h4-h5
>>>>f6-f7 h5-g5 f7-e8 g5-f5 e8-e7
>>>>25. Score: 1.8; Time: 3; nodes: 75500; PV: a1-b2 a7-b6 b2-c3 b6-b7 c3-d2 b7-c7
>>>>d2-d3 c7-b6 d3-e3 b6-c7 e3-f3 c7-d7 f3-g3 d7-e7 g3-h4 e7-f6 h4-h5 f6-f7 h5-g5
>>>>f7-e8 g5-f5 e8-e7 f5-e4 e7-d7 e4-f5
>>>>26. Score: 1.8; Time: 4; nodes: 83644; PV: a1-b2 a7-b6 b2-c3 b6-b7 c3-d2 b7-c7
>>>>d2-d3 c7-b6 d3-e3 b6-c7 e3-f3 c7-d7 f3-g3 d7-e7 g3-h4 e7-f6 h4-h5 f6-f7 h5-g5
>>>>f7-e8 g5-f5 e8-e7 f5-e4 e7-d7 e4-f5 d7-e7
>>>>I play a1-b2
>>>>Nodes: 83k; NPS: 21414; QNodes: 0; MO%: 98; FHA: 2; HC%: 55; Hash%: 58; NM%: 0;
>>>>egtb hits: 0; Ext's: 0; Total time searched: 3.906
>>>>
>>>>(MO% = first-move fail-high percent; FHA isn't important; HC is the percent of
>>>>time the hash move causes a cutoff; Hash% is the percent of time I find the move
>>>>in the hash table; the rest is un-important)
>>>>
>>>>(told you it was a long post)
This page took 0.01 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.