Author: Robert Hyatt
Date: 05:25:37 11/07/97
Go up one level in this thread
On November 07, 1997 at 05:17:25, Kai Lübke wrote: >On November 06, 1997 at 08:01:44, Bruce Moreland wrote: > >>On November 06, 1997 at 07:25:01, Kai Lübke wrote: >> >>>Have all programs play with permanent brain turned _off_. >> >>By turning this off you mess up the time allocation system of the >>programs. They are designed to by run with this feature on, and they >>may have code that attempts to increase the liklihood that thinking will >>be done on the opponent's clock. > >Does that imply that it is _in general_ flawed to test two programs >against each other with PB off? (I do this quite often since I don't >always have two machines available) > >--- >Shep Hard to say. When Ed/I played the NPS match, where crafty had 100x the time rebel had, we did no thinking on opponent's time. We found that we did some important things there that were not getting done. I've since done some tuning to eliminate part of this, but am not certain about whether it is totally reliable or not. IE I used to allocate extra time up front "knowing" I would make it up with correct predictions later. But with ponder=off, I never did and got into time trouble. Ed had a similar problem, and whether he fixed it or not I don't know. It is still better than letting both run 100% of the time as a well-behaved program can get skunked by one that is a cpu-hog.
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.