On Mon, 8 Sep 2003, SP LEE wrote:
> At move 140 of the attached game, gnugo overlooked the only move to kill
> white at B7. The following pattern solves the problem. I will run
regression
> test later.
Certainly worth trying. It would be more logical to put it in
owl_vital_attackpats.db, however, as otherwise gnugo will continue to
count the space to the left of (*) as a secure eye when it should not.
Also, below is a suggestion to make the constraint more precise.
Arend
> +Pattern A722
> +#spl added 3.5.2
> +
> +XXO?
> +.*XO destroy eye, valid both on center and edge
> +XX.O
> +
> +:8,s,value(45)
> +
> +XXO?
> +.*XO
> +AA.O
> +
> +;lib(A)==3
How about:
BBO?
c*XO
AA.O
; (lib(A) == 3 || lib(BB) == 2) && proper_eye(c)