Re: [gnugo-devel] New owl attack pattern

[SP LEE]

> SP Lee wrote:
> > My concern about the owl vital attack patterns is that they are only
> > matched once during owl analysis. So I would only add patterns which
are
> > more urgent, such as a serious threat, or destroying two eyes at
same
> > time to them. This pattern could be formed after several steps in
owl
> > analysis.
>
> No, the vital patterns are matched at every owl node.
>
> > In my test, the owl attack pattern still gets tried both in my
> > test case and in your test case.
>
> This depends on the exact circumstances but the vital patterns are
> always matched. Ordinary patterns are shortcut if the owl code is
> sufficiently sure of two eyes or is getting too deep.
>
> /Gunnar
>

I happened to find that the pattern A719 in owl_attackpats.db does the
similar thing, only it can't be matched on the edge. So it seems to be
reasonable to modify A719 instead of adding a new pattern. However, I'm
again not familiar with the constraints in A719 and I'd like to here
some comments about it. At first glance, we may just want to merge the
constraints together, but I'm not sure some of them is really necessary.

SP Lee