[ProgClub list] ai-checkers in jj5-test
jj5 at progclub.org
Tue Oct 18 08:50:48 AEDT 2011
On 18/10/2011 7:13 AM, John Elliot wrote:
> Coming at this from a different direction, given the state of the board
> that we're starting with, there are 9 red pieces and 7 white pieces for
> a total of 16 pieces on the board. Any piece could be located on one of
> 33 different locations, being any valid square (32) or 'off the board'
> (1). A piece can also be either normal or a king, which is an extra bit.
> So there are 34 states to model for each piece, being whether it is a
> king or not, whether it is on the board or not, and where on the board
> it is. You can represent 34 states in 6 bits, which you could pack into
> 8 bits having a 16 byte data structure for the location/state of each
> piece, or a 6*16 = 96 bits = 12 byte data structure to model pieces and
> their location.
Just for the record, my explanation there was wrong, but it's not really
worth going in to why, because that method of modelling the game states
isn't as good as the other method.
(Essentially I 'added one' when I should have 'timesed by two'. Or
something like that. :)
More information about the list