Author: Matthew White
Date: 14:06:29 03/14/03
Go up one level in this thread
On March 14, 2003 at 03:51:18, Albert Bertilsson wrote: >Hi Matt! > >Thanx for the reply, when you say it sounds so natural. Of course more knowledge >will result in more nodes search since ther will be many new interesting >variations for the engine to look at. > >I though of the eval as the guide to a single target, so making it better would >make the nodes searched drop. But of course the eval is a guide to MULTIPLE >target positions so more eval makes the engine realize there are new positions >to search because they have advantages that wasn't noticed before. > >I really hope though that adding even more knowledge will not make the node >count go up much more, because then it will be very expensive to add more. > >/Regards Albert You are very welcome. I had trouble with the concept myself. For example, I wanted to increase an engine's ability to stack rooks on open files. It ended up causing the engine to move a rook to an open file any time it couldn't find anything to do. An important thing to remember is balancing the eval parameters. A little change goes a long way. Read all you can about positional chess. Matt
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.