Author: Kurt Utzinger
Date: 21:28:52 09/14/02
Go up one level in this thread
On September 15, 2002 at 00:19:20, John Merlino wrote: >On September 14, 2002 at 17:14:41, Kurt Utzinger wrote: > >>Zugeordnete Felder 1.g4 - Cheron 0724 >>[D]8/8/K6p/6p1/8/8/k5PP/8 w - - 0 1 >> >>In this position the old TheKing 3.12d under Fritz7-GUI finds the winning move >>1.g4 within 7s. CM9 with the same amount of hashtables (32 MB) needs 3m7s, but >>CM9 with only 4 MB hashtables is the fastest with 3s. A real mystery for me, all >>on P4 1800/512. >> >>Analysis by The King 3.12d: Fritz7-GUI (32 MB hash) >> >>1.Kb5 Kb2 2.Kc5 Kc2 3.Kd5 Kd2 4.Ke5 Ke2 5.Kf6 g4 6.Kf5 g3 7.hxg3 Kf2 8.g4 Kg3 >>9.Ke4 Kxg2 10.Kf5 Kf3 >> = (0.13) Depth: 3 00:00:00 >> = (0.09) Depth: 20 00:00:04 1512kN >>1.g4 Kb2 2.Kb6 Kb3 3.Kb5 Kc3 4.Kc5 Kd3 5.Kd5 Ke3 6.Ke5 Kf2 7.Kf6 Kf3 8.Kf5 Ke3 >>9.h3 Kd4 10.Kg6 Ke5 11.Kxh6 Kf6 12.Kh5 Kg7 13.Kxg5 >> +- (1.47) Depth: 20 00:00:07 2232kN >> +- (4.49) Depth: 24 00:02:49 41876kN >> >>(Utzinger, MyTown 14.09.2002) >>******************************************************************************** >>TheKing 3.23 (CM9 GUI) 32 MB hash >> >>Time Depth Score Positions Moves >>0:04 17/21 0.08 1772968 1.Kb5 Kb2 2.Kc5 Kc2 3.Kd5 Kd2 4.Ke5 >> Ke2 5.Kf6 g4 6.Kf5 g3 7.hxg3 Kf2 >> 8.Kf4 Kxg2 9.g4 Kf2 10.Ke4 Kg2 >> 11.Kf5 Kf3 >>0:09 18/22 0.08 2783014 1.Kb5 Kb2 2.Kc5 Kc2 3.Kd5 Kd2 4.Ke5 >> Ke2 5.Kf6 g4 6.Kf5 g3 7.hxg3 Kf2 >> 8.Kf4 Kxg2 9.g4 Kf2 10.Ke4 Kg2 >> 11.Kf5 Kf3 >>3:07 18/22 2.19 39768632 1.g4 Kb3 2.Kb5 Kc3 3.Kc5 Kd3 4.Kd5 >> Ke3 5.Ke5 Kf2 6.Kf6 Kf3 7.Kf5 Kg2 >> 8.Kg6 Kf3 9.h3 Kg3 10.Kxh6 Kh4 >> 11.Kg6 Kxh3 12.Kxg5 >>---- ---- ---- ---- ---- >>******************************************************************************** >>TheKing 3.23 (CM9 GUI) 4 MB hash >> >>Time Depth Score Positions Moves >>0:00 12/16 0.10 249552 1.Kb5 Kb3 2.g4 Kc3 3.Kc5 Kd3 4.Kd5 >> Ke3 5.Ke5 Kf3 6.Kf5 Kg2 7.Kg6 Kh3 >> 8.Kh5 Kxh2 9.Kxh6 >>0:00 13/17 0.10 350152 1.Kb5 Kb3 2.g4 Kc3 3.Kc5 Kd3 4.Kd5 >> Ke3 5.Ke5 Kf3 6.Kf5 Kg2 7.Kg6 Kh3 >> 8.Kh5 Kxh2 9.Kxh6 >>0:01 14/18 0.31 587468 1.Kb5 Kb2 2.Kc5 Kc2 3.Kd5 Kd2 4.Ke5 >> Ke2 5.Kf6 g4 6.Kf5 g3 7.hxg3 Kf2 >> 8.g4 Kxg2 9.Kf4 Kh3 >>0:03 14/18 1.51 960806 1.g4 Kb3 2.Kb5 Kc3 3.Kc5 Kd3 4.Kd5 >> Ke3 5.Ke5 Kf3 6.Kf5 Kg2 7.Kg6 Kf3 >> 8.h3 Ke4 9.Kxh6 Kf4 10.Kg6 >>0:04 15/19 1.49 1158272 1.g4 Kb3 2.Kb5 Kc3 3.Kc5 Kd3 4.Kd5 >> Ke3 5.Ke5 Kf2 6.Kf6 Ke3 7.Kg6 Kf4 >> 8.h3 Ke5 9.Kxh6 Kf6 10.Kh5 >>0:47 16/20 2.19 9617559 1.g4 Kb3 2.Kb5 Kc3 3.Kc5 Kd3 4.Kd5 >> Ke3 5.Ke5 Kf2 6.Kf6 Ke3 7.Kg6 Kf4 >> 8.h3 Kg3 9.Kxh6 Kh4 10.Kg6 Kxh3 >> 11.Kxg5 >>0:51 17/21 2.19 10532568 1.g4 Kb3 2.Kb5 Kc3 3.Kc5 Kd3 4.Kd5 >> Ke3 5.Ke5 Kf2 6.Kf6 Ke3 7.Kg6 Kf4 >> 8.h3 Kg3 9.Kxh6 Kh4 10.Kg6 Kxh3 >> 11.Kxg5 >>1:15 18/22 3.48 15574649 1.g4 Kb3 2.Kb5 Kc3 3.Kc5 Kd3 4.Kd5 >> Ke3 5.Ke5 Kf2 6.Kf6 Ke3 7.Kg6 Kf4 >> 8.h3 Ke5 9.Kg7 Ke6 10.Kxh6 Kf6 >> 11.Kh5 Kg7 12.Kxg5 >>1:17 19/23 3.48 16035289 1.g4 Kb3 2.Kb5 Kc3 3.Kc5 Kd3 4.Kd5 >> Ke3 5.Ke5 Kf2 6.Kf6 Ke3 7.Kg6 Kf4 >> 8.h3 Ke5 9.Kg7 Ke6 10.Kxh6 Kf6 >> 11.Kh5 Kg7 12.Kxg5 >>1:20 20/24 3.81 16776827 1.g4 Kb3 2.Kb5 Kc3 3.Kc5 Kd3 4.Kd5 >> Ke3 5.Ke5 Kf2 6.Kf6 Ke3 7.Kg6 Kf4 >> 8.h3 Ke5 9.Kxh6 Kf6 10.Kh5 Kg7 >> 11.Kxg5 Kf7 12.h4 Kg7 13.Kf5 >>******************************************************************************** > >Johan was gracious enough to answer my e-mail on a Saturday (although it was >probably very early on a Sunday for him!). Here is what he said: > >-------------- >Basically, the shape of the search tree (hence the size) depends very much on >"luck". Different programs, different versions, and different settings (even the >hash table size or k-best mode) may change the "solution time" by a factor 10 or >100 easily. > >Chess programmers are of course not very happy with this behaviour, but the >average performance (ie scoring points) always goes first. In other words, >theoretical correctness is always sacrificed in favor of average program speed >and development time, with every programmer using his favourite trick(s) to keep >the damage small. Another example of the luck factor can be observed in the >thread CCC:251808. >-------------- > >There you go! > >jm Johan de Koning's explanations sound logical. Nevertheless I have the feeling that something could be wrong with the way TheKing 3.23 is handling such positions. I am going to search for further examples. Regards Kurt
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.