Re: [gnugo-devel] very specific defense pattern

[Paul Pogonyshev]

SP LEE wrote:
> >+
> >+Pattern Def70
> >+# pp New Pattern (3.5.1)
> >+
> >+OOOO*         a very specific position (defend the stone on the first line)
> >+O....
> >+XO...
> >+-----
> >+
> >+:8,D
> ><< specific-position.sgf >>
>
> I thought this pattern is little strange first. Then I understand the
> opponent might capture the stone on the first line if lacking of liberty.
> Should this be an endgame pattern? Should the liberty check be added? If
> there is no enemy stone above *, this defense may not be urgent.

Evan wrote:
> I believe that we are discussing a tactical defense pattern.  These
> patterns are rather special, in that they are only checked at stackp == 0,
> and are only tried in cases where they are needed (ie, defense patterns
> are only tried for potentiall attackable strings, and vice versa).  So, it
> makes sense to try this defense any time the 1st line stone is attackable
> and no other times; a constraint to enforce this is unneeded because of
> the way the patterns get used.  Of course, I could be misremembering here,
> it's been a while since I looked at that code.

Evan is right.  you may look at function defense_callback() in `worm.c'.
it checks whether a string is attackable and whether the proposed move
does indeed defend it.  so, there is no need to duplicate it in the
constraint.  note that unlike in many other databases, most patterns
in `defense.db' don't have constraints.

Paul