Author: Stuart Cracraft
Date: 20:40:58 03/06/06
Go up one level in this thread
On March 06, 2006 at 23:36:22, Dann Corbit wrote: >On March 06, 2006 at 22:14:27, Nathan Thom wrote: > >>>>3. Search inefficiency (branching factor of a good program is definitely under >>>>4) >>> >>> * My branching factor is about 2-3 for these kinds of positions. >> >>How are branching factors calculated? I get wildly different values at each ply >>as each side usually has different numbers of moves available to them... and at >>the root node, its always the full number of moves isnt it? >> >>e.g, for 8/6k1/6Pp/3r1P2/6K1/n3BP2/1p6/4R3 w - - 3 51 >>I get branching factors at each ply of 26 2 20 4 16 3 13 3 10 > >The simplest and most accurate way to determine your branching factor is to >divide the time to complete iteration N+1 by the time to complete iteration N >(don't bother computing it if you had an interrupt halt calculations -- >calculate it only if it finished naturally). That's what I do, then I average them all together for the current iterative deepening 1-N set for the given search. After that I average all those averages together across a test suite to get the final branching factor. The former are br= in my listing and the latter are bf= which is an ongoing average of the averages. Stuart
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.