Author: Robert Hyatt
Date: 05:25:22 10/22/98
Go up one level in this thread
On October 22, 1998 at 03:34:50, Ernst A. Heinz wrote: >On October 21, 1998 at 16:33:13, Robert Hyatt wrote: >> >>Alpha/beta only proves that non-best moves are non-best, no matter which >variant >>is used... and in the case here, any attempt to extract more information >(which >>is best, second-best, etc...) comes at enormous cost... > >Okay, now I see where our "disagreement" springs from. I interpreted your >statements as relating to the distinct root *moves* whereas your point of >view was from the root *position*. Of course, putting any *position* into >any variant of alpha-beta gives you no more information than the best move >and its score. > >However, if you refrain from doing minimal-window searches for any of the >root moves (i.e., the resulting ply-1 *positions*) you can calculate >accurate scores for all of them. But this surely breaks the alpha-beta >paradigm at ply 0 = the root position. > >=Ernst= OK.. back on same page. :) In effect it adds W*2*sqrt(W^(D-1)) nodes... since we now do W full alpha beta searches, but each to a depth of D-1... I do this in "annotating" a game where a user asks for the best 6 moves at each point in the game. But it costs substantial time.
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.