Author: Uri Blass
Date: 12:44:49 10/21/02
Go up one level in this thread
On October 21, 2002 at 15:27:36, Peter Berger wrote: >On October 21, 2002 at 15:16:25, Uri Blass wrote: > >>On October 21, 2002 at 14:36:00, Peter Berger wrote: >> >>>On October 21, 2002 at 14:31:46, Uri Blass wrote: >>> >>>>On October 21, 2002 at 14:28:23, Peter Berger wrote: >>>> >>>>>On October 21, 2002 at 12:15:23, Kurt Utzinger wrote: >>>>> >>>>>>a) >>>>>>since at SSDF they are testing at 120'/40 [many thanks for this great work] but >>>>>>95% of all users are (only) playing blitz games >>>>>> >>>>>An interesting statement. I wonder if it is true that 95% of the users prefer >>>>>blitz games, wouldn't it be more useful if the SSDF tested the programs at a >>>>>faster time control? Although it is true that slow time controls improve the >>>>>quality of the games, it might be a bad idea to test at time controls only a >>>>>tiny group of people use. >>>> >>>> >>>>I think that a lot of users are interested in slow time >>>>control and they may not use computers for comp-comp games. >>>> >>>>I believe that the correspondence players are interested in the >>>>program that is best in slow time control. >>>> >>>>They do not use programs for comp-comp tournament but for analyzing >>>>their games. >>>> >>>>Uri >>> >>>I agree - but for this group the SSDF results might be of little interest, too, >>>because they are much too fast. >>> >>>Here results in complicated testsuites for long time searches might be much more >>>interesting. >> >>Test suites are not a good substitute for games. >> >>I believe that the best estimate that we can get for >>the ability of programs at long time control is the ssdf list. >> >>Uri > >Again I agree - but you said that people were interesting in analyzing their >games or to help them in correspondence games. > >If this doesn't mean that they want to let their programs play all the moves ( >which looks a little senseless), we talk about doublechecking moves. No I believe that programs can be strong also in finding good positional moves. > >Then results in testsuites might be indeed be more helpful than games at 3 >minute/move time control. > >To take it a little further : why is 40/120 such a special time control for >computergames ? > >This makes no sense at all to me, if this isn't the time control people use most >often with their programs. There is a reason. With slower time control the number of games that you can get is too small for significant results. 120/40 is the best compromise maybe 240/40 could be better but it is clear that you cannot get significant result in 24 hours per move. 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.