Author: Will Singleton
Date: 19:32:43 12/21/98
Go up one level in this thread
On December 21, 1998 at 21:16:31, Chris Moreton wrote: > >I can't decide whether I'm concerned or encouraged now by Rival's poor blitz >performance. It searches about 50,000 nodes per second which I had assumed was >very slow because it is completely Object-Oriented and therefore attributed its >poor performance to this. Judging from the other figures posted here I think >maybe it has nothing to do with speed but rather just that it's not very good! > >Chris Chris, I've noticed that performance is very sensitive to small changes in my program. Very minor bugs or design errors can have a large effect. For example, I changed the repetition code recently to "fix" something, and also made some other changes. The program immediately tanked on ICC, although the wac scores were not affected. I found that I'd put the rep test in the wrong place, cutting off good moves. When I fixed it, the program started playing better again. So, I suggest analyzing your online games, finding the move or moves that led to problems, and then monkey with the code to get a better move. Not a quick process, but analyzing online games (as opposed to test suites) has helped me. Will
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.