Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Measuring closeness to a minimal tree

Author: Vincent Diepeveen

Date: 07:08:21 04/06/03

Go up one level in this thread


On April 06, 2003 at 10:00:36, Dan Andersson wrote:

> How could you ever get a measure of the minimal graph if you disregard
>transpositions? A-B withot memory is just a curio. One way of looking at
>transpositions is as compression.

I disagree Dan.

All compression i know compresses lineair. Hashtables would compress now
exponential which is pretty funny form of compression.

We can of course mention that hashtables can cause 3 effects and personally i
don't care in favour of which you are
  a) hashtables improve branching factor
  b) hashtables effectively make the number of average moves less in a position
  c) hashtables shorten the search depth needed to get to depth n.

In DIEP around 1996-1997 extensive experiments conducted with DIEP, my average
branching factor was 4.2 with hashtables. Also i counted the number of average
moves in a position (not counting positions where king is in check as those get
extended anyway) which was to my big surprise nearly exactly 40. And not 35 as
theory indicates (note that counting the incheck positions would bring down that
40 back to 39.xx or so).

It's easy to calculate what an average number of moves = 40 causes for minimal
branching factor to get 1 ply deeper when not using hashtables:
  sqrt 40 = 6.3

Now my 4.2 was not a minimal branching factor but *practical* branching factor.
Most likely that 6.3 is in reality 9.0

Best regards,
Vincent

>MvH Dan Andersson



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.