gnugo-devel
[Top][All Lists]
Advanced

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

Re: [gnugo-devel] fast liberties & patterns anchored in goal


From: Gunnar Farneback
Subject: Re: [gnugo-devel] fast liberties & patterns anchored in goal
Date: Wed, 06 Feb 2002 19:27:42 +0100
User-agent: EMH/1.14.1 SEMI/1.14.3 (Ushinoya) FLIM/1.14.2 (Yagi-Nishiguchi) APEL/10.3 Emacs/20.7 (sparc-sun-solaris2.7) (with unibyte mode)

Trevor wrote:
> The speed improvement by forcing the anchor to be in the goal are
> really big, and it might be worth considering doing the same with
> the owl patterns.

I've considered doing this before, but never gotten around to it. I'm
certain it would be a big win for the owl patterns. One should be
aware though, that some patterns are useful regardless which of the
stones is included in the dragon, such as the O stones in this
pattern: 

| Pattern D1334
| # This is not handled as a lunch because capturing only gives a false eye.
| 
| XOX          break out into the center, followup
| OXO
| o*o
| 
| :|,-,value(95)
| 
| XOX
| OXO
| o*o
| 
| ;owl_escape_value(*)>0

This can be solved by splitting the pattern in two and changing the
symmetry to 8 in the asymmetric one.

> The main pattern diagram now accepts two new characters: P & Y
> (P follows O, Y follows X).  P is the same as O, but that this
> point must be the anchor.  Likewise, for Y & X.

I recommend changing P to Q since it is visually more similar to O.
Actually mkpat already has an interpretation of Q, but this has long
been abandoned and can safely be discarded. When it was used, in a
failed escape route estimation experiment, it was equivalent to an O
but marked for special treatment in the callback.

/Gunnar



reply via email to

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