Author: David Dory
Date: 22:44:30 02/09/03
Go up one level in this thread
On February 09, 2003 at 18:10:44, Divy wrote: >Thanks for the response, Ive been asking these questions for years with no luck. >It just makes sense to me to tweak these engines. The parameters (of some) are >far too vast to not take notice. While the human "eye" probably couldnt notice >a +/-10 point difference in king safety, everyone here knows it makes a >difference to the computers. So how did the programmers come up with the engine >"defaults"? Incrementally changing values one by one, and then conducting engine >matches? Seems too tedius. This just seemed like the logical place to ask such >questions. Forgive me for sounding like a bore for those in the "know". > >Regards >Divy In the case of Chess Tiger, Christophe has written that he has written an extensive automated series of tests which he uses heavily. He believes, as you have written, that humans are unlikely to guess which settings are the best without such extensive tests. It's important to remember that many authors have been refining their chess settings for several years. From experience, they know (or believe they know), which setting will be most likely appropriate, and will test them first, against a selected number of other likely settings, and problem test sets. A few settings that do well at this stage, then play against other programs and/or humans. From a more objective POV, Christophe's automatic test method must be admired. From an experiental Point Of View, Hyatt's approach of having Crafty play game after game on the net, has certainly allowed him to fine tune many evaluation settings. I can't recall a word being said about how Fritz is tuned. Dave
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.