Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: methods for testing changes

Author: Brian Richardson

Date: 07:10:43 05/30/01

Go up one level in this thread


I think testing is another example where science and art must be combined in
chess programs.  Currently, I also have a rather haphazard system.  There are a
handful of test coded directly in Tinker that I have been using for more than a
year.  They are from different game phases and give me some idea if a change has
dramatically broken something or not, but only because I have a "feel" for
Tinker.  I then run a potential new version thru WAC to look at the mix of
positions solved and the total time.  Sometimes less is more here.  Finally, I
just put the new version up on ICC and let it play for at least a few days.
When I have time, I would like to implement temporal difference learning, but
the results so far seem mixed (Exchess?).

Brian



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.