Author: Uri Blass
Date: 04:50:25 04/26/01
Go up one level in this thread
On April 26, 2001 at 07:31:42, Vincent Diepeveen wrote: >On April 26, 2001 at 02:37:39, Uri Blass wrote: > >>game 1: >> >>In this position >>R7/3r2k1/4b1p1/P2N3p/5P1P/3K1BP1/5b2/8 w - - 0 1 >> >>Deep Fritz played 56.Kc4 0.47/15 Rc7 57.Kb4 0.28/16 Rd7 and the following >>position happened >> >> >>[D]R7/3r2k1/4b1p1/P2N3p/1K3P1P/5BP1/5b2/8 w - - 0 1 >> >>Deep Fritz played 58.Nb6 0.09/16 time when it pondered 8:57 >>and used 1007 knodes per second. >> >>I believe that it is a mistake in the time allocation of Deep Fritz. >>It should use more time for 58.Nb6 and the time of pondering was not enough. >> >>The following facts are obvious: >> >>1)Deep Fritz had a possible move that force repetition(58.Kc4) >>2)The score of Deep Fritz dropped and was only slightly above 0 >> >>I believe that programs should use more time in these situations. >>The time control was 40/90 and it means that Deep Fritz probably had more than >>45 minutes to calculate and in this situation that is not common I would prefer >>to use at least half of them. > >Oh this is not a matter of time division, i think Fritz time division >is more than ok. It's about evaluation here. I agree that Deep Fritz evaluation was wrong but deeper search could probably avoid the error because Deep Fritz's evaluation was negative in the next move. It could suspect that something is wrong in the evaluation because the score was positive and went down. It could also repeat the same position and I think that both of these factors should convince it to use more time in the relevant position(you cannot get evaluation that is always right and if you have possible repetition and a reason to suspect that the evaluation is wrong then using more time is a good idea). Uri
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.