Re: [gnugo-devel] New owl attack pattern

[Gunnar Farneback]

SP Lee wrote:
> Thanks for the comments. How about the following modified version of
> this patch?

Unfortunately it breaks strategy3:122. 

> I think it's important to get the pattern get matched first,
> then let the owl analysis to check the possible follow-up results. So I
> didn't add more contraints to it.

One should remember though that the owl code has a limited amount of
nodes and a limited branching factor, so one should avoid generating
ineffective moves if possible (with a reasonable effort).

Below is an alternative patch which makes a minor revision of the
constraint in A719 and introduces an edge variation of the pattern.
It solves the original problem in this discussion and has no effects
on the regressions.

/Gunnar

Index: patterns/owl_attackpats.db
===================================================================
RCS file: /cvsroot/gnugo/gnugo/patterns/owl_attackpats.db,v
retrieving revision 1.94
diff -u -r1.94 owl_attackpats.db
--- patterns/owl_attackpats.db  21 Aug 2003 10:45:52 -0000      1.94
+++ patterns/owl_attackpats.db  14 Oct 2003 17:49:23 -0000
@@ -2712,6 +2712,7 @@
 
 Pattern A719
 # gf New pattern. (3.3.20)
+# gf Revised constraint. (3.5.2)
 
 ?XXO   reduce eye space 
 X.*X
@@ -2723,7 +2724,23 @@
 Xa*X
 ?XXb
 
-;owl_mineye(*)==1 && xlib(b)==2 && !attack(c) && !oplay_attack(*,a,b,b)
+;owl_mineye(*)==1 && xlib(b)==2 && !oplay_attack(*,c) && !oplay_attack(*,a,b,b)
+
+
+Pattern A719b
+# gf New pattern. (3.3.20)
+
+|XXO   reduce eye space 
+|.*X
+|XX.
+
+:8,s,value(50)
+
+|XXc
+|a*X
+|XXb
+
+;xlib(b)==2 && !oplay_attack(*,c) && !oplay_attack(*,a,b,b)
 
 
 Pattern A720