Re: On game space size

[Avijit Ghosh <aghosh@xxxxxxxxxxxxxxxxx>]

On Wed, 28 Apr 1999, Imrich Kovac wrote:

> Generating of configurations:
> For each place on  board pick randomly from (-1,0,1) 
> (black,empty,white).
> May be that is the same as yours scheme 1?


	If you do it this way aren't you are sampling a 
gaussian distribution around w,b,e around 33% w 33% b 33% e? 
However I would guess that there should be a bias towards
empty boards.. i.e. there are many more legal board positions
when there are more empty settings than when there are less.
So if you sample using a uniform distribution on w,b,e I would guess
it will be harder to hit the board space w/ empty boards thus making
the % legal boards / total boards lower than it should be??

	Out of curiousity do people ever use stochastic search
methods for AI searching? I took an undergrad class ages ago,
which went on and on about A* and B and all sorts of lexiographic
search methods names :) that basically had to do w/ pruning trees and
ordering  and whatnot.. but instead of doing it that way (if you have a
big space) could it be easier to use something like MC methods..? 


	-me