RE: FW: On game space size

[Avijit Ghosh <aghosh@xxxxxxxxxxxxxxxxx>]

On Wed, 28 Apr 1999, Pieter Cuijpers wrote:

> (Is this clear ???)
> 	As clear as mud. :)
> 	I don't know much about Monte Carlo. Is it kind of sampling? 
> 	But I will be very interested in the final result of your research.
> 
> Monte Carlo is a nice name for a dirty method.
> The Monte Carlo method generates random samples of points in a search 
> space and based on the evaluation of those points it tries to tell
> something about the whole space.

I have some probably naive questions as well :) 

How are you generating your configurations?? ie how do you know
how many pieces to put on of each color? Is it all random from scratch
i.e. pick X random blacks Y random white's put generate X random
coordinates for black Y random coordinates for white, place on board,
evaluate for legality, repeat?

Or do 'grow' your board? w/ resets? i.e randomly place a black piece
(evaluate) randomly place white piece (evaluate), repeat, all the while 
counting and dropping all configurations which are illegal, until the
board gets filled up, and then repeat.

Also are you using an acceptance/rejection criterea at all for
test configurations? i.e. how would the numbers be different if you
biassed your sampling w/ a move evaluation function from a
generic go playing program?  (Using scheme 2)

> 
> verifying whether a board situation is legal can be done by flooding
> the liberties over all the white stones. (so white stones with
> neighbouring liberties are taken of the board) If there are any white
> stones left they must have no liberties and therefore the board 
> situation is illegal. Do the same for black  and we know whether a
> board situation is legal.

	do you have to actually flood the board or just check to see
if there are pieces on the board that have no liberties? 

	Thanks :)
	-avi