Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Branching factor, make me confuse more that ever.

Author: Andreas Stabel

Date: 06:39:34 04/05/00

Go up one level in this thread


On April 05, 2000 at 00:57:48, Tom Kerrigan wrote:

>This is just a semantic argument. What you call stuff depends on how you look at
>it.
>
>1) If you think a chess program searches the chess game tree, then the tree's
>branching factor will be 38, although only a tiny fraction of that tree is
>actually visited.
>
>2) If you consider the chess program's search tree separate from the chess game
>tree, then the branching factor is 3 (or whatever).
>
>I think it's pretty shortsighted to delcare one viewpoint correct and the other
>incorrect. It just depends on who you ask. But I think we both agree that (2) is
>more common. Might as well use it.
>
>-Tom

Perhaps you missed it, but here is some statistics I have from parsing Dann
Corbits game collection. I posted it in this thread earlier.

Read the five files A.PGN, B.PGN, C.PGN, D.PGN and E.PGN.
The files had a total of 25458158 lines and 945950820 bytes
   giving an average of 37 bytes per line.

Read a total of 1496327 games with 112004996 moves including variations,
giving an average of 8.45 bytes per move.
There were 6605 variations (left parentesises) containing 63879 moves.

Excluding the variations there were 111941117 moves giving rise to
  113437444 positions, including the start position, with a total
  of 3545011178 legal moves giving an average of 31.25 legal moves in
  an average position.

The 111941117 moves was:
Pawn moves     : (27.15%) 30386622
Knight moves   : (18.23%) 20403265
Bishop moves   : (15.86%) 17748466
Rook moves     : (18.24%) 20418343 (including castlings)
Queen moves    : (11.97%) 13403989
King moves     : (10.89%) 12190285 (including castlings)
Short castlings: ( 2.09%)  2337283
Long castlings : ( 0.24%)   272570

Checking moves     : ( 4.98%)  5571618 (including mating moves)
Mating moves       : ( 0.04%)    48748
Stalemating moves  : ( 0.001%)     986
Hit moves          : (20.65%) 23116992 (including en passant moves)
En passant moves   : ( 0.08%)    86830
Pawn two moves     : ( 7.59%)  8494513
Promotion to queen : ( 0.07%)    82298
Promotion to rook  : ( 0.0008%)    839
Promotion to bishop: ( 0.0003%)    281
Promotion to knight: ( 0.001%)    1519

Results of the 1496327 games were:
White won: (38.70%) 579064
Drawn    : (29.89%) 447291
Black won: (29.11%) 435569
Unknown  : ( 2.30%)  34403

Decided games:
White mated     : ( 1.42%) 21259
White stalemated: ( 0.03%)   479
Black mated     : ( 1.84%) 27489
Black stalemated: ( 0.03%)   507

So by this evidence the branching factor could be said to be 31.
Perhaps som food for thought :)

Regards
Andreas Stabel



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.