[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [gnugo-devel] Proposed owl_determine_life() improvement

From: bump
Subject: Re: [gnugo-devel] Proposed owl_determine_life() improvement
Date: Sun, 3 Nov 2002 20:16:24 -0800

> Thank you, it sure will provide some good examples. But I see that
> you're actively working on this issue. Maybe I'd better concentrate
> on something else... Anyway, here's a comment.

No, I think it's good if you want to look at these examples. I
doubt if I work much more on this if you want to do something about it.

> Mixing the eyes arithmetic with escape potential looks a bit odd to
> me. I think it will possibly cause problems in do_owl_attack() and
> do_owl_defend(). I guess this could be the reason of some of the
> failures you are currently getting with this patch. Are you positive
> this is the right way to treat the problem ?

I'm *not* sure this is the best approach.

But I wouldn't characterize the patch this way. 

The eye arithmetic here just says that if the worm dies then the
group is alive, hedging the bet a little if it consists of only a
single worm. It is through the parameters min, probable and max that
sniff_lunch reports its findings. To give these values of 2 is just
saying alive, not really eye arithmetic.

> Another possible way I was thinking about would be to try to refine
> the job of owl_determine_life() by trying to find out if a move that
> captures a lunch also creates some potential for another eye.
> For instance in nngs4:30

The function owl_substantial tries to estimate this. It is
a very expensive function and as written only valid when stackp==0.


reply via email to

[Prev in Thread] Current Thread [Next in Thread]