Re: [gnugo-devel] EJ45

[Evan Berggren Daniel]

The issue is that as written, the pattern matches in nngs4:450.  The
position is:

|.....
|.*..X
|.XOOX
|..XXO

The two stones are in trouble, and globally should probably be sacrificed.
A play at *, where EJ45 matches, is absolutely awful.  One option is to
simply make it not match at the edge;  however, edge blocks like that can
be valuable.  Maybe it should be a separate pattern?

I assume the original intent of the pattern was to preserve center
influence against an opponent hane; there should be a constraint to make
sure this is what is happening.  Would an omoyo() constraint work?  I
don't know the influence code workings very well.

Anyway, what I settled on was the constraint I gave, that made sure the
play was reasonable tactically.  Maybe widening the pattern makes it
unneeded, but I think not having some sort of constraint there is risky.

Thanks

Evan Daniel

 On Thu, 6 Mar 2003 address@hidden wrote:

>
> Evan wrote:
>
> > This patch solves nngs4:450 (in my previous email).  EJ45 is matching at
> > B5, where it shouldn't.  This patch expands EJ45 slightly to fix this.
> > The constraint may still need some work.
>
> The constraint is:
>
> +; !safe_omove(a) && !oplay_attack(*,B) && !oplay_attack_either(*,c,*,B)
>
> The first oplay_attack seems rather unlikely. How likely is it that
> X can capture B after O plays *?
>
> I don't see the motivation for the second oplay either.
>
> Dan
>
>
> _______________________________________________
> gnugo-devel mailing list
> address@hidden
> http://mail.gnu.org/mailman/listinfo/gnugo-devel
>
>