Re: How does the computer tell a live group from a dead group

[David Fotland <fotland@xxxxxxxxxxxxxxx>]

Hi Sam,

The problem has not been solved completely.  It's very hard to
tell when a group is alive, or if there is a fight, which will win,
or that some complex seki exists.  Every go program does it a
little differently.

I read tactics for groups with 4 or fewer liberties, then read
connections and eye diagonals.  I patern match eye shapes, count
effective liberties, and apply some heuristics for group
strength.  This usually gives the right answer, but not always.

Once you knwo which gropus are alive, finding territory at
the end of the game is trivial.  Just remove dead stones and
do a flood fill.

It always seemed like a hard problem to me.

David Fotland

At 02:21 PM 2/25/99 -0500, Sam Sloan wrote:
>Here is a problem I worked on years ago, and never solved it to my
>satisfaction.
>
>Obviously it has been solved, but I would like to know the solution.
>
>How does the computer tell a live group from a dead group?
>
>Next, how does the computer count the score accurately at the end of the
game?
>
>Obviously, there must be several algorithms for this, but I would like to
>know one.
>
>To those of you who have never actually tried to solve this problem, let me
>tell you that this is not as trivial a problem as it may seem.
>
>Sam Sloan
>
>
>