Re: [gnugo-devel] endgame pattern

[Evan Berggren Daniel]

On Tue, 12 Aug 2003, Paul Pogonyshev wrote:

> this pattern solves the new test endgame:890

> +Pattern EE108
> +# pp New pattern (3.5.1)
> +
> +?OX?     very small endgame move
> +O.*.
> +----
> +
> +:8,OXes,value(0.5)
> +
> +?OX?
> +Oa*b
> +----
> +
> +; proper_eye(b) && oplay_attack(*,b,a,a) && oplay_defend(a,*,a)

This looks wrong to me, for a couple reasons.  First, in endgame:890, the
value of the play is 2/3 in gnugo counting, not 1/2.  That is to say it is
worth less than this:

XOOO
X*XO
XOOO

and the same as this:

XXOO
X*XO
XXOO

Second, this doesn't account for the fact that the play could be played as
a larger ko if the outer situation were different.  That is, on the
endgame:890 board, w T4, b T3, and then if things were different, w cuts
at S3 and fights ko.  In that case, the play could be worth a lot more.
Potentially, it could be ko for the life of a large group.

Evan Daniel