Author: Andreas Guettinger
Date: 01:33:48 11/27/02
Go up one level in this thread
On November 26, 2002 at 21:48:28, David Rasmussen wrote: >I found the bug... > >time nodes depth score pv >00:00:00.00 984n 01/02/17 -0.321 19. Nxf6+ Nxf6 >00:00:00.03 1063n 01/02/17 --------------------------------------------------- >00:00:00.06 1962n 02/03/12 -1.418 19. Nxf6+ Nxf6 20. Nd2 Nxg4 > 21. Bxg4 Bxg4 >00:00:00.06 2302n 02/03/16 -1.231 19. cxd5 axb3 20. dxc6 bxc2 > 21. cxd7 Nxd7 >00:00:00.08 2493n 02/03/16 --------------------------------------------------- >00:00:00.09 3774n 03/04/19 -1.231 19. cxd5 axb3 20. dxc6 bxc2 > 21. cxd7 Nxd7 >00:00:00.12 5599n 03/04/19 --------------------------------------------------- >00:00:00.12 8727n 04/05/19 -1.231 19. cxd5 axb3 20. dxc6 bxc2 > 21. cxd7 Nxd7 >00:00:00.17 16kn 04/06/19 --------------------------------------------------- >00:00:00.19 23kn 05/07/19 -1.231 19. cxd5 axb3 20. dxc6 bxc2 > 21. cxd7 Nxd7 >00:00:00.22 34kn 05/07/19 -0.898 19. Nxf6+! >00:00:00.25 45kn 05/08/18 +0.010 19. Nxf6+ gxf6 20. Nd2 h5 > 21. cxd5 Qxd5 >00:00:00.29 55kn 05/08/20 --------------------------------------------------- >00:00:00.39 94kn 06/09/20 -0.049 19. Nxf6+ Nxf6 20. Nxc5 d4 > 21. Nxe6 Qxe6 22. Bf4 Nxg4 > 23. Bxg4 Qxg4+ >00:00:00.56 146kn 06/09/23 --------------------------------------------------- >00:00:00.72 216kn 07/11/23 +0.084 19. Nxf6+ Nxf6 20. Nxc5 d4 > 21. Bxd4 Bxc5 22. Bg2 Qc8 > 23. Bxc5 Qxc5 24. Bxb7 >00:00:01.03 326kn 07/11/23 --------------------------------------------------- >00:00:01.37 480kn 08/14/23 +0.084 19. Nxf6+ Nxf6 20. Nxc5 d4 > 21. Bxd4 Bxc5 22. Bg2 Qc8 > 23. Bxc5 Qxc5 24. Bxb7 >00:00:01.65 581kn 08/14/23 --------------------------------------------------- >00:00:02.39 941kn 09/16/24 +0.084 19. Nxf6+ Nxf6 20. Nxc5 d4 > 21. Bxd4 Bxc5 22. Bg2 Qb6 > 23. Bxc5 Qxc5 24. Bxb7 >00:00:03.11 1235kn 09/16/24 --------------------------------------------------- >00:00:05.78 2495kn 10/16/24 -0.032 19. Nxf6+ Nxf6 20. Nxc5 dxc4 > 21. Nxe6 Qxe6 22. Qxc4 Bxe3 > 23. Qxe6 Bxf2+ 24. Kxf2 fxe6 >00:00:07.03 3025kn 10/16/25 --------------------------------------------------- >00:00:12.70 5776kn 11/21/26 -0.018 19. Nxf6+ Nxf6 20. Nxc5 Nxg4 > 21. cxd5 Qxd5 22. Bxg4 Bxg4 > 23. Rfc1 Bf5 24. Qc3 Rae8 >00:00:15.69 6991kn 11/21/26 --------------------------------------------------- >00:00:24.62 11Mn 12/22/29 -0.122 19. Nxf6+ Nxf6 20. Nxc5 Nxg4 > 21. cxd5 Qxd5 22. Bxg4 Bxg4 > 23. Ne4 Bd4 24. Bxd4 Qxd4 > 25. Rae1 Rae8 >00:00:29.94 13Mn 12/22/29 17/42 19. Nd6 > > Time used: 00:00:30.00 > nps: 456882 nodes: 13711044 Q-nodes: 5348044 (39%) > move order: 88% > hash probes: 2528374 hash hits: 735380 (29%) hash usage: 97% > pawn hash probes: 10680903 pawn hash hits: 10472532 (98%) > EGTB probes: 0 EGTB hits: 0 (0%) > >I bet you're curious to know what the bug was :) > >I can't explain it exactly, because that would require that you have detailed >knowledge about the way my functions work. But there was a _huge_ bug in the way >that qsearch looped through moves. Basically, in certain situations, if there >were n captures generated, qsearch could end up trying only the expected best >one of them, n times! That is, doing the exact same search n times! > >What I don't understand is how Chezzz was at all able to play chess with this >sort of bug, let alone get a decent rating on ICC. I guess it shows that qsearch >can be pretty inaccurate or even buggy, and still "work". >Wow... > >There's noting better than when you've just fixed a bug :) > >/David Maybe it plays twice as strong as before, now. :) regards Andreas
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.