Author: Uri Blass
Date: 11:01:46 04/14/03
Go up one level in this thread
On April 14, 2003 at 09:52:43, Jouni Uski wrote: >At http://www.computerschach.de/freeware/index.htm. > >Frank is genius: he has final solution to problem: how to count rating from test >positions! Only thing, which need change is that doubling gives today >about 50 points - not 70. (Speedup from K6 to Athlon in SSDF is 3.7 not >2.7 as in calculations). > >Jouni I think that it is dependent on the time control and if you use fast time control than doubling gives more than 50 points. Another point is that you cannot ignore the fact that in most cases if you compare programs with a big difference of more than 200 elo(For example Fritz8 and Nejmet3.06) then the program that is better in tactics has also better evaluation. I expect Fritz8 to beat Nejmet3.06 even if you give Nejmet time advantage to do it equal to Fritz8 in test suites. The point is that a better evaluation may make the program only 10 elo better in tactical test suites thanks to better order of moves but 50 elo better in games. 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.