Author: Ed Schröder
Date: 03:34:22 10/25/04
Go up one level in this thread
On October 25, 2004 at 04:12:42, ThatsIt wrote: >>Hi Gerhard, >> >>Can you share your fiindings in a nutshell? >> >>Ed > >Hi Ed ! > >I will try. > >I've made the following: >engine A vs engine B match with 3 repetitions >under the same conditions (reboot, same startpositions, >same order of loading the progs, same timecontrol, >clearing learnfiles and so on...) >timecontroll 5+1 under CB-S8-GUI > >junior8 vs fritz8 (both nativ !) >match no. 1 = 8.5 - 11.5 >(reboot....) >repetition no. 1 = 7.5 - 12.5 >(reboot....) >repetition no. 2 = 7.0 - 13.0 >(reboot....) >repetition no. 3 = 5.5 - 14.5 > >then the same procedure: > >junior8 vs hiarcs9 (both nativ !) >match no. 1 = 12.5 - 7.5 >(reboot....) >repetition no. 1 = 7.0 - 13.0 >(reboot....) >repetition no. 2 = 10.0 - 10.0 >(reboot....) >repetition no. 3 = 7.0 - 13.0 I understand but the error margin is way too high for a final conclusion. Let me explain the problem with fast time controls like these. The average move-time is about 5-6 seconds, the decision to go one ply deeper (or not) around 2-3 seconds. Now Windows is famous for interrupting your PC every x seconds snooping following its own unfathomable ways, doing some maintenance etc. This cost processor time eating tenths of seconds just enough to prevent engine X to go to the next iteration and so a different game is born because of the randomness of Windows. >then the same procedure with to other engines, >but timecontroll 15+1: This is already a better criterion as the above mentioned effect will occur less. >ruffian 2.1.0 vs shredder 8 >match no. 1 = 9.0 - 11.0 >(reboot....) >repetition no. 1 = 6.5 - 13.5 >(reboot....) >repetition no. 2 = 5.0 - 15.0 >(reboot....) >repetition no. 3 = 8.0 - 12.0 > > >after starting a match absolute nothing has done with the machine, >no mousemoving, no look at the interim score ... >no progs were running in the background. > >as you can see i've done all possible to give them the same >conditions. Yes. >funny, insn't it ? Yes, but still explainable, therefore I used 800 games in my experiment. Cherio, Ed
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.