Author: Dieter Buerssner
Date: 10:45:30 09/05/02
Go up one level in this thread
On September 05, 2002 at 10:47:32, Robert Hyatt wrote: >I don't think our serial searches are very bad. IE I get the best move >first 92% of the time. I'm not sure how much farther I can go with that >as there will _always_ be flaws that only a deep search exposes, when you >sort moves in some arbitrary way. I guess you meant the fraction of beta cutoffs in the first move you try, by the 92%. Then, this number may also be misleading. Is it really the best move, or just any move, that cutoffs? Many more moves may actually cutoff, but usually we don't know this (unless writing some experimental unefficient minimax code for collecting the statistics). Other moves may cutoff much faster (with a smaller tree following). In the extreme, an alternative move may cutoff immediately from the HTs. Enhanced transposition cutoff checks for this, but in general, I think there are no well known algorithms to find the fastest cutoff move. I did some experiments for collecting some statistics a while back. IIRC with random move ordering, I often got close to 50% cutoffs in the first tried move, in the nodes that got a beta cutoff. Still, the search efficiency became (not surprising at all) extremely bad. Regards, Dieter
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.