Re: [gnugo-devel] look at regress file

[Evan Berggren Daniel]

Given the classification of the problem as OWL_TUNING, and the fact that
the life of the white group is in question, the problem is that Gnu Go
needs to see L9 as an owl attack move.  Presumably it doesn't currently.
My guess is that this problem is actually very hard for Gnu Go.  The
reason that w must play L9 is that it is sente for life and expands the
eyespace; the reason it is sente is that it threatens to attack the black
group on the side.  Getting the owl code to see all this may be rather
difficult.  Currently, the owl code fails to find L9, and also fails to
see that L9 is a valid defense (owl_does_defend L9 J8 returns 0).  In
other words, the problem is more complex than it seems at first, I think.

Evan Daniel

On Fri, 28 Mar 2003, kevin yong wrote:

>
> Hi:
>   i m reviewing regression case:
>       http://www.public32.com/regress/?13x13:4
>   the problem is gnugo should answer L9, rather
> than J11.
>   i m trying to look inside the problem to see if
> anything i can do to make some improvement on it.
>   first of all, L9 never been raised by any
> move-generator. The Q is: giving all move_reasons
> listed in move_reasons.h, which reason(s) should we
> use to present L9 is a better move than J11?
>   could anyone have a quick look at the case and give
> some comments? thank you in advance.
> kevin.
>
>
> ______________________________________________________________________
> Post your free ad now! http://personals.yahoo.ca
>
>
> _______________________________________________
> gnugo-devel mailing list
> address@hidden
> http://mail.gnu.org/mailman/listinfo/gnugo-devel
>
>