Re: [gnugo-devel] owl tuning

[Evan Berggren Daniel]

On Fri, 6 Dec 2002, Arend Bayer wrote:

>
> Evan wrote:
>
> > Two new owl patterns, together they solve 13x13:73.
>
> > Index: patterns/owl_attackpats.db
> > ===================================================================
> > RCS file: /cvsroot/gnugo/gnugo/patterns/owl_attackpats.db,v
> > retrieving revision 1.66
> > diff -u -r1.66 owl_attackpats.db
> > --- patterns/owl_attackpats.db      12 Nov 2002 13:41:55 -0000      1.66
> > +++ patterns/owl_attackpats.db      25 Nov 2002 14:41:50 -0000
> > @@ -1702,6 +1702,25 @@
> >  ;(owl_escape_value(b) > 0)
> >
> >
> > +Pattern A424
> > +#evand new pattern (3.1.13)
> > +
> > +O..X
> > +o.*.
> > +..x.
> > +????
> > +????
> > +
> > +:8,-,value(35)
> > +
> > +O..X
> > +o.*a
> > +..cb
> > +????
> > +????
> > +
> > +;(x_somewhere(c) && !oplay_attack_either(*,a,b,*,b))
> > +;|| (!x_somewhere(c) && xplay_attack_either(c,*,a,b,*,b))
>
> Did you mean
> ;|| (!x_somewhere(c) && !xplay_attack_either(c,*,a,b,*,b))
> here?
>
> I am not sure whether the constraing might be too expensive.

I actually meant it as written.  Here's why:

If there is already an x stone there, and we can cut anyway, do so.
If there isn't an x stone there, and x playing there would prevent us from
cutting, block before it's too late.

It feels somewhat messy, but I think it's right.  As for expense, I
assumed it would be cheaper to check this than spend an owl node
unneccessarily.

Perhaps the best way to solve both problems is with two patterns, one with
the stone and one without.

>
>
> > +Pattern D1386
> > +#evand new pattern (3.3.13)
> > +
> > +x.Oo       keima connection to escape
> > +x..O
> > +x...
> > +..*.
> > +?...
> > +????
> > +
> > +:8,E,value(65)
> > +
> > +#x.Oo
> > +#x..O
> > +#x...
> > +#..*.
> > +#a...
> > +#bcde
> > +
> > +#;(owl_escape_value(a) > 0) || (owl_escape_value(b) > 0) ||
> > +#;(owl_escape_value(c) > 0) || (owl_escape_value(d) > 0) ||
> > +#;(owl_escape_value(d) > 0)
> That's certainly     ^^^^ "e" here.

Oops.

> Did you run regression?

I actually don't remember.  I certainly ran some of them.  I'll rerun them
now.

Thanks

Evan Daniel