Author: Stuart Cracraft
Date: 14:38:30 08/11/04
Go up one level in this thread
On August 10, 2004 at 06:03:35, Daniel Clausen wrote: >On August 09, 2004 at 14:34:48, Stuart Cracraft wrote: > >[snip] > >>I will not improve a program based on theory alone and games with me. >>I am too poor a player to go that route. > >Did you consider adding xboard/winboard support and let it play automatically >against other engines on your local machine or on a chess server like ICC? > It still comes down to interpreting a game. I am not a good enough player to do that accurately. >I think there are many things which you can try on a test-suite with good >results, especially at the beginning of your engines life. (like SEE _will be_ >better than MVV/LVA - if implemented/used correctly) But afterwards it gets much >harder to prove that a certain change is worth it. (esp. in your eval - IMHO) > >Sargon I think that is right. As a program grows, test suites get worse and worse at measuring slighter and slighter differences. They're great for the first few weeks or months of a program, but after that the return is less and less. Stuart
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.