Author: Robert Hyatt
Date: 18:58:16 08/03/99
Go up one level in this thread
On August 03, 1999 at 19:06:48, Dave Gomboc wrote: >On August 03, 1999 at 14:44:31, Robert Hyatt wrote: > >>put it on ICC. That once per year might become once per day. Which will >>seem much too high to accept, IMHO. Look at my "easy move" code. It is >>_very_ restrictive. And even it has blown out a couple of times that I saw, >>and who knows how many more times that I didn't... > >Ed said that his code was simple, but that doesn't imply permissive. It could >be that his implementation is restricted much in the same fashion that yours is, >or for that matter, in a totally different, but also effective manner. > I have no idea what he does. The only context I have is that he mentioned doing a 1 ply search without alpha/beta at the root to get real scores for each move. It is easy enough to test whether this works well or not. ICC is a fiery cauldron waiting to expose any 'hole'. :) >Let's face it: the program is going to make mistakes some of the time, no matter >what. It doesn't matter whether an "easy move heuristic" fails once a year or >once a day, as long as the program plays better overall. In a previous >conversation some months ago, Ed said he suspected that his easy-move heuristic >costed 5 elo points at most, and that satisfying his customers was well worth >it. Some of us wondered whether it was even a net gain, elo-wise. > >I think it's quite possible to get a net gain out of an easy-move heurstic. >Obviously you do, too, or you wouldn't have the one that is in Crafty. > >Dave I only do it because it was requested so many times. I eliminated it in Cray Blitz 15 years ago... Same sort of reason as Ed, actually...
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.