Author: Sven Reichard
Date: 14:34:38 09/29/99
Go up one level in this thread
On September 29, 1999 at 09:24:55, Jari Huikari wrote: >My CM3000 gives an approximation of humans rating according to how >he/she can find best moves in series of positions. > >How good approximation for a program strenght this kind of method would be? > > > Jari To be consistent, the set of positions would have to be constant over some time; and if the set is rather small, it is no problem to tune a program to do well on them. This approach is taken in the Bratko-Kopec test, originally a very small sample, but now a huge collection of positions with well-defined best moves. I'm not sure they return a value in terms of ELO.
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.