Author: David Rasmussen
Date: 07:04:34 11/25/02
I've seen Christophe in the past making the point that testing methodology is what becomes important, once you've implemented all the trivial and near-trivial stuff. So, can we discuss testing methodologies? How many can we list? What are they good for? Christophe, can you give some hints perhaps? Just to name a few obvious ones: Self-play Match-play Online-play Test-suites etc. What are these good for, and not so good for, respectively? And how about time controls? Apart from specifically testing the engines worth in bullet games, can bullet games be used for testing anything useful? In test-suites, when is nps important, when is nodes per full ply or nodes to solution important etc. ? Please come up with more questions yourselves. I find that the most difficult thing to test is a change of the evaluation function or a change of the pruning/extensions scheme. /David
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.