Author: Robert Hyatt
Date: 18:10:12 01/30/01
Go up one level in this thread
On January 30, 2001 at 17:46:28, Severi Salminen wrote: >Hi! > >What is the definition of BF and how is it calculated (this has probably been >discussed N times, but...)? Now I'm calculating like this: > ______________________ >BF=(DEPTH-1)|(Nodes/root_move_count) > >So it means (DEPTH-1)th root. > >Example (from Vincent's position): DEPTH=9 plies, Nodes=2'000'000 and we have 38 >legal moves in root. > ____________ >BF=8|2'000'000/38 = 3.9 > >Could this be close to the truth? > >Severi There are two terms that are intermixed often: 1. "branching factor" usually means the average number of branches from a node, which for chess averages 35-38 (depending on who you believe) for the whole game. 2. "effective branching factor" is roughly calculated as you mentioned. IE nodes for iteration N-1 divided into nodes for iteration N. Or you can use the time. This is really "effective branching factor" because there is _nothing_ that can reduce the true branching factor other than pure forward pruning, where some moves are thrown out at each and every ply with no searching of any kind. IE for Crafty, EBF is roughly 3.0... which is actually impossible on the surface since alpha/beta reduces the branching factor by roughly the square root of it. square root of 38 is just over 6, which is a reasonable branching factor for alpha/beta. To get lower, in the case of crafty, rather than culling some moves by not searching them at all, I simply search many of them to a lower than usual depth because of the null-move R value... numbers around 3 are very good IMHO. I only wish mine stayed there all the time. But it can easily run up to 10 or more in oddball positions where the move ordering breaks down at deeper depths...
This page took 0.01 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.