Author: Uri Blass
Date: 22:34:44 07/15/05
Go up one level in this thread
On July 16, 2005 at 00:38:19, Pallav Nawani wrote: >On July 15, 2005 at 13:44:01, Uri Blass wrote: > >>On July 15, 2005 at 12:45:04, Pallav Nawani wrote: >> >>> >>>>...and 40 moves in 40 minutes is too fast for reliability, IMHO, >>> >>>Wrong. All available data shows that for ratings, time control does not matter, >>>unless it is very small. Even bullet tournaments have highly similar ratings in >>>most cases. What really matters is consistent testing conditions and the number >>>of games played. >>> >>>>especially for >>>>those who use the software in correspondence chess. The 40 moves in 2 hours used >>>>by SSDF is best. >>> >>>SSDF tests on 1 Ghz Pcs , IIRC. CEGT tests on time control of 40/40 on 2 Ghz pc. >>>Therefore CEGT time control is equivalent to 1Hr20Min per 40 moves on SSDF >>>hardware which is quite ok. >>> >>>Pallav >> >>No >> >>SSDF test on 1.2 ghz and not on PIV that is slow for chess. >>CEGT test on something equivalent to 40/40 on PIV 2 ghz >Okay, I stand corrected. > >>CEGT also test without pondering so I think that CEGT time control is equivaelnt >>to something like 30 minutes/40 moves in ssdf conditions. > >You are overestimating the impact of pondering. Most of the time engines just >ponder on wrong moves, so that time is just wasted, even though incorrect >pondering will fill the hashtables with (sometimes) useful data. But savings in >case of a incorrect pondering are so small they are negligible. > >Pallav I think that engines ponder on correct moves near half of the cases and I guess pondering is equivalent to being 1.5 times faster. 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.