Author: Robert Hyatt
Date: 11:36:19 08/18/05
Go up one level in this thread
On August 18, 2005 at 14:29:50, Tord Romstad wrote: >On August 18, 2005 at 10:55:08, Robert Hyatt wrote: > >>That is interesting, because this past week, I have played 43 games (long time >>control). 21 vs shredder, 5 vs prodeo, 10 vs junior, 1 vs gnuchess, 2 vs fruit, >>and a couple I don't remember, all via my "crafty" ICC handle. Care to guess >>how many times crafty's king safety caused a problem? Absolutely zero. So I'm >>not sure I buy the "very often" part of the above. It certainly has some >>king-safety issues, as does most programs I see playing. But it is not a "very >>often" issue. Unless you force Crafty to play certain openings that it doesn't >>normally like to play... > >I confess that I do force Crafty to play openings it doesn't normally play. I >always run my tests from fixed starting positions, without opening >books. An opening book optimised for the program is certainly useful >in important tournaments, but I think using them in test matches is a >bad idea (unless, of course, the quality of the book is what you want to >test). When the program doesn't understand certain types of positions >very well, it is better to try to improve its understanding of such positions >than to pick opening lines which avoid them. > >Tord Right, but my point was this. If I don't test from positions where I don't think Crafty will do well, then someone that does test from those positions is going to draw a conclusion that might well be meaningless in a WCCC event where Crafty will never reach such positions.
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.