Re: [gnugo-devel] Patch: Strategic effect for strategic attacks and defenses

[Inge Wallin]

Gunnar wrote:

> Inge wrote:
> > trevorb:390 FAILs for other reason
> >               The chosen move, F7, has a valuation of 6 points for
> >               "defends against combination attack on several worms",
> >               which a) does only capture in ko and b) the correct
> >               move also protects against.  
> > 
> >               The problem here is that atari-atari only generates
> >               one defense point.
> 
> Wrong, atari_atari (or strictly speaking combinations()) generates up
> to four defense points. In this particular case F7, G7, and G6 (the
> correct move) are found.

This has to be new.  I didn't check the code, but took the "fact" from
my memory.  That's what you get when you are away for 6 months.  :-)

> But it's correct that atari_atari sometimes misses the best defense
> when there are lots of candidates. This patch solves the problem by
> forcing atari_atari to generate nearly all possible defense
> moves.
> 
> This is done by marking all potential defense moves during the
> atari_atari reading, i.e. moves in the attack sequence, liberties of
> moves in the attack sequence, and moves defending some threatened
> string in the sequence. To determine which of them are really
> effective, we play the move and see whether the combination attack has
> gone away.
> 
> The regression delta is good, 12 passes, no failures.

Yay!  Way to go.  Nice to see that the work I put into the tests paid
off. 

     -Inge