Re: [gnugo-devel] multiple either_move reasons

[Evan Berggren Daniel]

On Mon, 26 Aug 2002, Arend Bayer wrote:

> On Mon, 26 Aug 2002, Evan Berggren Daniel wrote:
>
> > /* FIXME: We need a policy if a move has several EITHER_MOVE
> >  * reasons. Most likely not all of them can be achieved.
> >
> > Would a reasonable answer here be to add the odd numbered highest values
> > together?  eg, if there are four possible groups to save, and not all of
> > them can be, the most valuable one is saved, and then the next most
> > valuable one is killed, and then the third is saved, etc?  Granted it's
> > not ideal, but wouldn't it be better than just taking the top move?
>
> Well, I think the true answer is that only look-ahead will be able to
> solve this. If we know the move attacks either A or B, and also either B
> or C, it could be that it can definitely capture 2 of the three, or that
> it could capture either B or (A and B)...we can't know without reading
> this out.

I see...  we're worried about one move attacking more than two groups.  I
agree, reading is the only really good solution.  Though perhaps it works
better to decide that there has to be a way to connect some of the groups
so that there are only two, and call it a connection problem?  I'm afraid
my understanding of go gets a little shaky here...

> At the moment we just add min(size(A), size(B)) + min(size(B), size(C)).
> This doesn't make much sense, and better rules could be applied here;
> they would have to be tried out -- feel free to try s.th..

OK, that makes more sense now.

> But I don't think it is a very frequent problem.

Agreed.

Thanks

Evan Daniel