Author: Kurt Utzinger
Date: 22:14:41 07/15/03
Go up one level in this thread
On July 15, 2003 at 18:24:13, George Tsavdaris wrote: >On July 15, 2003 at 17:47:23, Kurt Utzinger wrote: > >>On July 15, 2003 at 16:07:44, George Tsavdaris wrote: >> >>>On July 15, 2003 at 15:21:17, Robert Hyatt wrote: >>> >>>>Here is a position reached in a game on ICC between Crafty and an >>>>unnamed opponent: >>>> >>>>[D]8/2k2p2/7p/ppPK4/6P1/P6P/8/8 w 0 1 >>>> >>>>This seems like a simple "distant passed pawn" idea where black's king isss >>>>closer to the remaining pawns after the queen-side pawns are all gone. I'd >>>>hope not many would play c6 and lost instantly, but at least one commercial >>>>engine does. Goes to show that _anything_ can happen in a comp vs comp >>>>game of chess. :) >>>> >>>Deep Junior 8 makes ~4 minutes in a PIV 1500MHz to see c6 is losing and >>>it's the first choice, until it sees Ke5 or Ke4. So in time control of >>>120'/40+120'/40, in my computer, it would PROBABLY play the wrong move, >>>but in a fast Dual/Quad it wouldn't. >> >> Amazing ... on my P4 1.8/48 MB hash, Junior8 single CPU just >> requires 1 second to throw away the losing move c6. Below the >> corresponding analysis: >> >>Neue Partie >>8/2k2p2/7p/ppPK4/6P1/P6P/8/8 w - - 0 1 >> >>Analysis by Junior 8: >> >>1.c6 b4 2.axb4 axb4 3.Kc4 Kxc6 4.Kxb4 Kd7 >> ² (0.65) Tiefe: 3 00:00:00 >> ³ (-0.30) Tiefe: 9 00:00:00 4kN >>1.h4 a4 2.Kd4 Kc6 3.h5 f6 4.Ke4 Kxc5 5.Kf5 Kd5 6.Kxf6 Kd6 >> ² (0.66) Tiefe: 9 00:00:00 6kN >> = (-0.15) Tiefe: 18 00:00:01 1216kN, tb=43 >>1.c6 b4 2.axb4 axb4 3.Kc4 Kxc6 4.Kxb4 Kd5 5.Kc3 Ke4 6.Kc4 f6 >> = (-0.04) Tiefe: 18 00:00:01 1329kN, tb=63 >>1.Ke4 Kd7 2.Ke5 Kc6 3.Kd4 Kb7 >> = (0.00) Tiefe: 18 00:00:02 1688kN, tb=104 >> = (0.00) Tiefe: 30 00:03:51 133376kN, tb=107832 >> >>(Utzinger, MyTown 15.07.2003) > >This is more strange. Junior at infinite analysis mode makes indeed 4 minutes >to see it's losing and not play c6. But in game mode 120'/40+120'/40 for >example, it sees that c6 is bad in 52sec and plays Ke4 in 1:27. All these >in a PIV 1500 MHz with 64 Mib hash. >The analysis of Junior 8 seems different from Deep Junior 8 so i wonder why? >Maybe someone with Deep Junior 8 on a PIV 1.8 GHz can help giving the analysis. > >Here is the Deep Junior's 8 analysis: > >1.c5-c6 b5-b4 2.a3xb4 a5xb4 > ² (0.65) Depth: 3 00:00:00 >1.c5-c6 b5-b4 2.a3xb4 a5xb4 3.h3-h4 > ² (0.66) Depth: 6 00:00:00 >1.c5-c6 b5-b4 2.a3xb4 a5xb4 3.Kd5-c4 Kc7xc6 4.Kc4xb4 Kc6-d7 > ³ (-0.30) Depth: 9 00:00:00 4kN >1.h3-h4 f7-f6 2.c5-c6 b5-b4 3.a3xb4 a5xb4 4.h4-h5 > ² (0.66) Depth: 9 00:00:00 4kN >1.h3-h4 Kc7-d7 2.c5-c6+ Kd7-c7 3.Kd5-c5 b5-b4 4.a3xb4 a5xb4 5.Kc5xb4 Kc7xc6 >6.Kb4-c4 > ² (0.53) Depth: 12 00:00:00 32kN >1.h3-h4 f7-f6 2.h4-h5 Kc7-d7 3.c5-c6+ Kd7-c7 4.Kd5-c5 b5-b4 5.a3xb4 a5xb4 >6.Kc5-d5 b4-b3 > ² (0.39) Depth: 15 00:00:00 223kN >1.h3-h4 a5-a4 > = (-0.11) Depth: 18 00:00:02 1520kN, tb=68 >1.c5-c6 b5-b4 2.a3xb4 a5xb4 3.Kd5-c4 Kc7xc6 4.h3-h4 Kc6-d7 5.Kc4xb4 Kd7-c6 >6.Kb4-c4 > = (-0.08) Depth: 18 00:00:03 1520kN, tb=81 >1.c5-c6 b5-b4 2.a3xb4 a5xb4 3.Kd5-c4 Kc7xc6 4.Kc4xb4 Kc6-d5 5.Kb4-c3 h6-h5 >6.g4xh5 > = (0.00) Depth: 21 00:00:05 2563kN, tb=299 >1.c5-c6 b5-b4 2.a3xb4 a5xb4 3.Kd5-c4 Kc7xc6 4.Kc4xb4 Kc6-d5 5.Kb4-c3 h6-h5 >6.g4xh5 > = (0.00) Depth: 24 00:00:13 6486kN, tb=1661 >1.c5-c6 b5-b4 2.a3xb4 a5xb4 3.Kd5-c4 Kc7xc6 4.Kc4xb4 Kc6-d5 5.Kb4-c3 Kd5-e4 >6.Kc3-c4 Ke4-f3 7.Kc4-d5 Kf3-g3 8.Kd5-e4 Kg3xh3 9.Ke4-f5 > ³ (-0.30) Depth: 27 00:00:52 24995kN, tb=10284 >1.Kd5-e4 Kc7-c6 2.Ke4-d4 Kc6-d7 3.Kd4-e5 Kd7-c6 > = (0.00) Depth: 27 00:01:27 34195kN, tb=19092 This all is indeed very strange. The Dual Junior8 seems to analyse in a completely other way than the single CPU version. 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.