Re: [gnugo-devel] owl_attackpat tuning

[Arend Bayer]

Trevor Morris wrote:

> - Currently the goal does not include stones added during the owl reading.
> These should probably be added with a distinct non-zero value, so these
> new stones can be anchors but won't break existing owl assumptions about
> the goal.  Namely, if all goal stones are captured, it's the dragon's 
> considered
> captured.

Hmm, that doesn't agree with the comment above owl_update_goal() (and
I am pretty sure it works as described):

/* Add the stone just played to the goal dragon if same_dragon is
 * 2. We also add all stones belonging to the same generalized string
 * to the goal. If same_dragon is 1, we only add the stones if at
 * least one stone of the generalized string already was part of the
 * goal. If same_dragon is 0, we don't add any stones at all.
 */

(This is called for each move by do_owl_defend().)

So maybe in the case you encountered the same_dragon field of the
responsible pattern was incorrect? (Or didn't apply to the situation.)

Arend