Author: Matthias Gemuh
Date: 03:15:47 04/26/04
Go up one level in this thread
On April 26, 2004 at 03:39:48, Jouni Uski wrote: >There has been some discussion about professional test methods. At least to >find good test method should be the key to success? But I simple cannot imagine >better test method, than playing a lot of games against a lot of opponents with >different time controls! What else there can be?? We can only quess how many PCs >SMK has at home / at Chessbase in 24h/7d work... > >Of course playing against humans is still better, but with longer time controls >in practise impossible. > >Jouni Many users have a strange imagination about programmer's engine testing. They think it is an issue of counting games won, lost or drawn. A programmer's testing has mostly to do with gathering and analyzing statistics related with innumerable internal engine tricks and secrets, e.g. a) how frequently does the time management trick fail ? b) how frequently could SEE safely replace QSearch ? c) how high is the NullMove efficiency ? d) how risky is the new pruning trick ? e) ... etc Beta testers can't help with the above. After all this, tying it up and the testing overall playing strength from user's view (using beta testers) is less time consuming. /Matthias.
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.