Re: [gnugo-devel] nando_3_9.4

[Arend Bayer]

I wrote:
> On Mon, 16 Sep 2002, Portela Fernand wrote:
>
> > Purpose:
> >
> >   This patch extends the OWL code so that it becomes aware of tactical
> >   gains and losses while trying to attack or defend a dragon.
>
> I have looked at it a bit more. I've put up an updated version at
> http://www.uni-bonn.de/~uzsxtn/nando_3_9.4a. (Mostly insignificant
> changes:
> * Merge with CVS, in particular in mark_changed_dragon() (addressed a
>   FIXME there)
> * A few coding-style-match-surrounding type cleanups.
> * Do worry even about loosing a single stone. (I think it's worth it
>   once the problem described below is addressed.)
> )
That last change actually breakes a couple of tests. One thing is we
shouldn't worry about ko captures here. The other that I didn't realize
that this new policy can cost a lot of owl nodes, sometimes causing
erratic FAILs because of uncertain owl results. I have an idea what to
do about this, but for now we should go with Nando's more conservative
approach here.

(I had put up that patch mostly to avoid someone else would have to redo
the merge work which I had to do anyway. Probably shouldn't have
cluttered this with my own experimental changes.)

Arend