Author: José de Jesús García Ruvalcaba
Date: 03:44:20 11/03/01
Go up one level in this thread
On November 02, 2001 at 13:58:37, Christophe Theron wrote: > > >Sometimes I get an advice from a strong player, but human players cannot tell me >if my program has become stronger or not. > >If you ask a human player you'll just get a subjective answer because here we >are talking of 30-50 elo points improvement of an already very strong opponent. >That's not something that can be detected by playing a few games against it. > >Part of the work of chess programming is developping a way of making sure that a >change in the program is really an improvement. When you see a strong chess >program, you can be sure that its programmer has found a good way of testing the >improvements. > >Chess programmers who are not willing to invest a large part of their time >developping this testing method end up being unable to improve their programs. >Because when a program is already very strong, any change has equal chances to >make it slightly stronger or slightly weaker. There is no way to tell if a >change will work or not before you test it very extensively. > > > > Christophe Hi Christophe, I know you are a little secreive about your testing methodology, so I will not be surprised or upset if you do not answer a question I have. Do you think your testing method will work for games other than chess? José.
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.