Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: King and pawn game tuner for your program

Author: Robert Hyatt

Date: 15:05:57 06/27/02

Go up one level in this thread


On June 27, 2002 at 17:36:59, Ferdinand S. Mosca wrote:

>On June 27, 2002 at 17:14:16, Robert Hyatt wrote:
>
>>On June 27, 2002 at 13:38:32, Ferdinand S. Mosca wrote:
>>
>>>
>>>The king and pawn game is a very interesting exercise to program, is it
>>>difficult to perfect this ending?.
>>>
>>>How long will it take your favorite programs to see that 36. h5 is good for
>>>black.
>>>[D]8/k6p/8/2P2p2/2P1p3/7K/1P3P2/8 b - - 0 35
>>>sample program analysis on Athlon xp 1600+, 64MB hash
>>>Analysis by Gromit 3.8.2:
>>>
>>>36...h5 37.Kg3 Kb7 38.b3 Kc7 39.b4 Kc6 40.f4 Kc7 41.b5 Kb7 42.c6+ Kb6 43.c5+ Kc7
>>>44.Kh3 e3 45.Kg3 h4+ 46.Kg2 h3+ 47.Kf3
>>>  ±  (1.09)   Depth: 18   00:10:55
>>>36...h5
>>>  ±  (1.09)   Depth: 19   00:10:55
>>>36...h5
>>>  ±  (1.09)   Depth: 19   00:11:09
>>>
>>>more than 11 min, and yet the score still favors white.
>>>
>>>
>>>Analysis by Green Light Chess 2.18:
>>>
>>>36...h5 37.Kg3 Kb7 38.b4 Kc6 39.f4 Kc7 40.b5 Kd7 41.c6+ Kd6 42.Kf2 h4 43.Kg2 e3
>>>44.Kf3 h3 45.Kxe3 h2 46.c5+ Kxc5 47.c7 Kxb5 48.c8Q h1Q 49.Qxf5+
>>>  ±  (1.27)   Depth: 19   00:03:16  154355kN
>>>36...h5 37.Kg3 Kb7 38.b4 Kc6 39.f4 Kc7 40.b5 Kd7 41.c6+ Kd6 42.Kf2 h4 43.Kg2 e3
>>>44.Kf3 h3 45.Kxe3 h2 46.c5+ Kxc5 47.c7 Kxb5 48.c8Q h1Q 49.Qxf5+
>>>  ±  (1.27)   Depth: 19   00:03:17  155502kN
>>>
>>>still favors white.
>>>
>>>Regards,
>>>Dinan
>>
>>
>>takes Crafty 1.5 seconds to lock on to the move, 5 seconds for the score
>>to drop below 0.00 (good for black) and 7 seconds to have a -1.5 score for
>>black...
>>
>>
>>               12     1.11   3.41   1. ... Kb7 2. Kh4 h6 3. b4 Kc7 4. Kg3
>>                                    h5 5. b5 Kb7 6. c6+ Kb6 7. Kh4 Kc7
>>                                    8. Kxh5
>>               12     1.52     ++   1. ... h5!!
>>               12     1.65   0.86   1. ... h5 2. Kg3 Kb7 3. b4 Kc6 4. Kh3
>>                                    f4 5. Kg2 h4 6. Kh3 e3 7. Kg2 exf2
>>                                    8. Kxf2
>>         (2)   12->   1.66   0.86   1. ... h5 2. Kg3 Kb7 3. b4 Kc6 4. Kh3
>>                                    f4 5. Kg2 h4 6. Kh3 e3 7. Kg2 exf2
>>                                    8. Kxf2
>>               13     1.96     ++   1. ... h5!!
>>               13->   2.21   0.47   1. ... h5 2. Kg3 Kb7 3. b4 Kc6 4. Kh3
>>                                    f4 5. Kg2 h4 6. Kh3 e3 7. Kg2 exf2
>>                                    8. Kxf2
>>               14     2.93     ++   1. ... h5!!
>>               14     5.26  -0.28   1. ... h5 2. Kg3 Kb7 3. b4 Kc6 4. Kh3
>>                                    f4 5. Kg2 h4 6. b5+ Kxc5 7. Kg1 e3
>>                                    8. Kf1 exf2 9. Kxf2
>>               15     6.42     ++   1. ... h5!!
>>               15     7.63  -1.58   1. ... h5 2. Kg3 Kb7 3. b3 Kc7 4. c6
>>                                    Kd6 5. b4 Kxc6 6. c5 Kd5 7. b5 Kxc5
>>                                    8. Kh4 Kxb5 9. Kxh5
>>               15->   7.93  -1.58   1. ... h5 2. Kg3 Kb7 3. b3 Kc7 4. c6
>>                                    Kd6 5. b4 Kxc6 6. c5 Kd5 7. b5 Kxc5
>>                                    8. Kh4 Kxb5 9. Kxh5
>>               16    10.33  -1.87   1. ... h5 2. Kg3 Kb7 3. b3 Kc7 4. c6
>>                                    Kxc6 5. b4 Kd6 6. b5 Kc5 <HT>
>>               16->  10.92  -1.87   1. ... h5 2. Kg3 Kb7 3. b3 Kc7 4. c6
>>                                    Kxc6 5. b4 Kd6 6. b5 Kc5 <HT>
>>               17    16.47  -1.73   1. ... h5 2. Kg3 Kb7 3. b3 Kc7 4. c6
>>                                    Kxc6 5. b4 Kb7 6. b5 Kb6 7. Kh4 f4
>>                                    8. c5+ Kxb5 9. c6 Kxc6 10. Kxh5
>
>Hello Bob,
>
>How do you penalyze a doubled passer (c4)?, or is this a passer at all, and will
>get a passer bonus? I thought that this should not be considered a passer.
>
>Dinan


I evaluate them pawn-by-pawn.  No real penalty here because of the connected
passer pair on the b and c files...  Actually, in looking at the code, I only
evaluate the most advanced passer on each file for the critical scoring.

Each pawn gets some usual positional stuff added in,  But then the endgame-
specific code only looks at the most advanced passer on each file, since it is
the one that is most dangerous...



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.