Re: [gnugo-devel] OWL failure !?

[SP Lee]

>
> > 1/
> > In this final position ,black never plays 133 =K13 (or L13) despite
> the fact
> > that there is a lot of matching patterns.
> > (at least for K13).
> >
> > GG invoked with --T first shows  the H16 dragon as dead(green)  but
> OWL does
> > rescucite it .
> >
> > When we make a trace of OWL's thinkings , we get this variation
> >
> >  Pattern A1001b found at K14 with value 50
> >           Trying black K14. Escape = 2. Current stack: B:K13 W:M13
> B:M12
> > W:N13 B:K14  (variation 6)
> > Variation 6: ALIVE (escaped)
> >
> > this conclusion of course is wrong.
> >
> > 2/ If we consider the previous white move : 132 white M11 , we can
> help
> > GNUGo with this kind of pattern which makes it to play 132 M13
> >
> > Pattern EC221
> >
> > ?oOoo
> > Xo.o?
> > o*.oO
> > O.X??
> >
> > :8,C
> >
> > of course , this does not solve the OWL failure .
> > I would have a question about whether 132 should be correctly played
> if OWL
> > could solve 133 .
> >
> > If someone has a pattern solution for the 133 move , i would be
> interested
> > to know about it too.
> > Thanks
> > md.
> >
>
> To my understanding this is a problem of escape value calculation,
> rather than pattern matching. So it couldn't be really solved by
adding
> any pattern.
>
>
> SP Lee
>

Sorry, this is a well known problem of "dragon is cut and owl doesn't
know". It has nothing to do with escape value calculation. There are
also some similar tests in the regression tests.


SP Lee