Author: Ratko V Tomic
Date: 09:24:33 10/15/99
Go up one level in this thread
> But what's to say that this isn't the move being pondered already? > Any asymmetry in the evaluation function could cause this to happen. As long as the 2nd "best" is predicted less often than the "best" one, it would still diminish the ponder time gain. In case of two Rebels it could negate most of it. > I think it'd be interesting to try, but it would also be quite difficult > to program, since Alpha-Beta does not return the 2nd best move. It > only returns the best move, and says "all other moves are worse > than this move". That assumes perfect move search order at the top of the move list, which need not be true. After all, the best picks do change from iteration to iteration (otherwise you wouldn't need to go beyond the ply-1 iteration). As mentioned in the DB thread, some programs may purposefully take extra time to compute the "true" value of the top few root moves to obtain some estimate of the error margin or to use it as a short list of candidates for further evaluations in a different way.
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.