Author: Andreas Herrmann
Date: 09:40:37 08/15/03
Go up one level in this thread
On August 15, 2003 at 07:52:41, Uri Blass wrote: >On August 15, 2003 at 07:23:04, Andreas Herrmann wrote: > >>On August 15, 2003 at 01:50:53, Ed Schröder wrote: >> >>>On August 15, 2003 at 01:04:21, Federico Corigliano wrote: >>> >>>>What's the formula for calculating the (effective) branching factor in Rebel 12 >>>>(DOS)? >>>>I'm interesting in add it to my engine. >>>> >>>>Greetings, >>>>Federico >>> >>> >>>Frederico, >>> >>>The most easiest and fair rule I could think off is to maintain the time of the >>>first mainvariation of the current iteration (t1) and the time of the first >>>mainvariation of the previous iteration (t2), then the branch factor is: >>> >>>bf=t1/t2 >>> >>>My best, >>> >>>Ed >> >>The branching factor descibes the average number of child nodes. So the formula >>should be >>bf = (t1-t2)/t2 >>or better >>bf = (n1-n2)/n2 >> >>where >>t1 = the total time for iterartion n >>t2 = the total time for iterartion n-1 >>n1 = the total nodes until iterartion n >>n2 = the total nodes until iterartion n-1 >> >>Andreas > >No >when people use the name branching factor they mean t1/t2(by your definition >the best possible branching factor is 0 and not 1). > >I do not know definitions from book but the important definition is definition >that people use and your definition is not what I see used. > >Uri > >Uri It's described in several papers. Just search for it with Google. Excample for the starting Pos, if you would search all nodes, then you have for white a branching factor of 20 and in the next iteration for black also a bf of 20. bf = (420-20)/20 = 20 {for depth 2) If you use the other formula, you got a wrong bf bf = 420/20 = 21 But black has not 21 answer moves to each move from white. And in a check mate position or in a stalemate position, where is no child node you have a bf = 0. Here some sample definitions: a) The average number of successor states for nodes in the search space is known as the branching factor. http://www.cee.hw.ac.uk/~alison/ai3notes/subsection2_6_3_4.html b) ...Because the branching factor (the number of children a node can have) in a binary heap is 2, the base of the logarithm in O(lg n) is 2 as well. ... http://www.csee.umbc.edu/courses/undergraduate/341/fall02/projects/proj4/341-Fall02-p4.shtml c) .. We assume that each internal node has the same number of children; this number is known as the branching factor. ... http://www.ics.uci.edu/~eppstein/180a/970417.html d) The branching factor is the number of possible moves (children). http://www.cs.rpi.edu/~hollingd/ai/lectures/games.ppt and so on... You will find a lot of papers/sources in the net with similiar definitions. But i also know, that a lot of people calculate the bf wrong. regards Andreas
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.