Author: Russell Reagan
Date: 18:08:34 06/10/02
Go up one level in this thread
I think as long as you keep one aspect of your testing constant when testing new versions, then you will get good results. For example, if you test a new version of movei against an old version of movei, and it performs better, then it's probably better. The next time around, you should probably test it against BOTH previous versions to make sure that your results are improving against ALL previous versions of the program. I suppose it's possible for the version 2 to be better than version 1, and then have version 3 beat version 2 regularly but lose to version 1. It's not likely, but it could happen. As far as testing it against other engine's, I think this is a good idea also. Of course, Tiger is very strong, so maybe we're wrong, but I think it can't hurt to test against other engine's. I think what you should do is test it against the same engine's for different versions. For example, if you played a match between movei 1 and GNU Chess 4 originally, then when you start testing movei 2, don't test it against GNU Chess 5 (or whatever the newest version is), test it against version 4, so that you know if it's improved over the same engines. You can of course test many many other programs against each version of movei, which would be a good thing, but I think you can only compare results of the new program against results of the old version when your program played a match against the SAME opponent. Russell
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.