Author: Robert Hyatt
Date: 07:46:06 06/13/04
Go up one level in this thread
On June 13, 2004 at 07:46:01, Franz Hagra wrote: >rating WM-Test = base 2600 + (2 x LQ) - [5 x (GZ : sum of positions) > >LQ= percentage of solved positions >GZ= total solve time (if a position is not solved the max. single test time >20min is used) > >some samples > >with dL=average solve time of solved positions; >St=solve time (including penalty time of 20 min for unsolved positions) >St= LQ*dL + (100-LQ)*20 > >Engine Fritz X3D on AMD TB 1400: > >LQ = 73 , dL = 2,27 min, St= 705 RATING 2711 Elo > >on this basis some thought experiments: > >LQ = 73 , dL=2,27/2, St= 623 RATING 2715 Elo > >that means that half dL will effect a growth of 4 Elo > >next we say the 73 position where found in 1 sec > >LQ= 73 , dL=1 sec, St= 542 RATING 2719 > >What more to say about the used formula? > >Hagra Nothing, other than to remember it is just one of 1,000 formulas that can be created to take the above data, and produce the above rating, for the above program. Won't have anything to do with _my_ program's rating, however... Using positions to predict rating is so far beyond flawed...
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.