[futurebasic] Re: [FB] tic-tac-toe (but really XFB)

Message: < previous - next > : Reply : Subscribe : Cleanse
Home   : May 1999 : Group Archive : Group : All Groups

From: "Phil Yates" <pyates@...>
Date: Mon, 10 May 1999 21:55:54 +0100
If any of you decide to dig out or write such a program, you might like to
turn it into an "expert system" using a learning algorithm, which is very
easy to implement. I believe it was called the "matchbox" learning method,
and I came across it originally in Martin Gardner's column in Scientific
American.

In tic-tac-toe you have a very small range of possibilities. And thus the
reason it was called the "matchbox" method. You had one matchbox
representing each square, and if you won, you "rewarded" each matchbox that
formed a winning play with a bead. And if you lost, you took a bead away.
To start, you looked for the box with the most beads, and played that
square. Thus the winning squares (boxes) had more beads in them, and the
losing squares less. You'll need one set of matchboxes for starting, and
another for responding.

Surprisingly, from no knowledge at all, the matchboxes would end up winning
or drawing with very few games played.

Doing this in FB is trivial once the basic rules have been entered. If you
decide to code TTT, you may like to add this feature.

Phil.

----------------------------------------------------------------------------
To estimate the amount of time it will take to produce system software,
estimate the time that it should take, divide by 2, then change the unit of
measure to the next highest unit.
----------------------------------------------------------------------------