Re: [gnugo-devel] FW: 9x9 Bug in 3.2 ? GnuGo passes after first move

[Evan Berggren Daniel]

On Sat, 12 Oct 2002, Gunnar Farneback wrote:

> Dan wrote:
> > Unfortunately, GNU Go does not see any way to make a live
> > group on the board. (Indeed, it is difficult for W to live.)
>
> I think we should force the engine to play moves even if all own
> stones look dead, under some additional condition like high handicap
> and/or a sufficiently small fraction of the board being filled.
>
> The easiest, although probably not the best, way to do this would be
> to revise all own stones as alive, rerun the main pattern matching and
> the move valuation, and disable certain safety and blunder tests.
>
> The wallyplus program should be useful for testing.

Just a random idea, feel free to ignore it if it isn't useful.

What if we assumed we get two moves in a row, and looked for useful ways
to do that?  Because if we believe that against proper play we can't make
life, but are going to try anyway, this is (I believe) equivalent to
assuming that our first move goes effectively unanswered, or is answered
so poorly as to be irrelevant.  So, perhaps an interesting idea would be
to make this assumtion explicitly.

Hope it's helpful.

Evan Daniel