Author: Frank Phillips
Date: 01:28:33 01/09/03
Go up one level in this thread
On January 08, 2003 at 23:01:08, Federico Corigliano wrote: >I will start to improve the strenght of my chess engines. I know that there are >situations where you add a new technique, a new knowledge to the evaluation >functions, and you don't know if it will result in an improve or all was a lost >of time. How do you test your program to see if the new change is good or bad? > >Regards > >Federico Corigliano I think this is a key question; and hope Ed Schroder will cover it on his 'programmer's stuff' web page. Test positions are helpful, but the results can often mispredict real play. For example you can add lots of tactical extensions to solve tactical tests quicker, but slow down the program in real games as it searches fruitless for tactical hits. Ultimately you test against your target - other computers or humans or both. I am still searching for a surrogate. You probably need less tactical but more positional strength against humans than computers for example. Even titled players fall prey to short horizon tactics. But I guess everyone with a program on ICC has been a victim of the blocked centre, slow king side pawn storm, followed by a piece sacrifice and mate. Frank
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.