Author: Ronald de Man
Date: 14:24:10 12/17/05
Go up one level in this thread
On December 15, 2005 at 15:55:03, Andrew Wagner wrote: >First, imagine that a perfect SEE algorithm existed. That is, given any >position, and a side to move, you could decide which piece to move without doing >any searching, and get it right 100% of the time. An algorithm that tells you *which piece* is involved in the best move? Not a realistic assumption for general positions, even using imagination. (And as others have remarked, this has nothing to do with the SEE algorithm.) >Now your search tree is down >to a branch with an ammortized size of about 12 branches per position (maximum >of 27 moves for the queen, 14 for rook, 13 for bishop, 8 for knight, 8 for king, >3 for pawn) with an average case in the middle game much lower. You still have >to search those 12 branches or whatever, but you're always only searching moves >for a single piece per node. This means you cut the branching factor by a factor of 2.5 or so. Assuming pure alpha-beta and perfect move ordering, you go from 2*30^(d/2) to 2*12^(d/2) nodes to reach depth d. With the same number of nodes, depth reached increases by a factor of log(30)/log(12) = 1.37.
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.