gnugo-devel
[Top][All Lists]
Advanced

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

[gnugo-devel] new defense pattern


From: Portela Fernand
Subject: [gnugo-devel] new defense pattern
Date: Mon, 23 Sep 2002 02:58:32 +0200

Hi all,

I've been toying with patterns and found this one to be apparently a bit
effective :

Index: patterns/owl_defendpats.db
===================================================================
RCS file: /cvsroot/gnugo/gnugo/patterns/owl_defendpats.db,v
retrieving revision 1.66
diff -u -r1.66 owl_defendpats.db
--- patterns/owl_defendpats.db  29 Aug 2002 10:43:35 -0000      1.66
+++ patterns/owl_defendpats.db  23 Sep 2002 00:27:57 -0000
@@ -2783,6 +2783,16 @@
 ;owl_proper_eye(a)


+Pattern D721
+# nn New pattern experiment (3.3.9)
+
+OoO        good shape can help
+o.O
+O*X
+
+:8,-,value(55)
+
+
 #########################################################
 #                                                       #
 #           Make eyeshape in the corner                 #


It looks like a very basic one. I thought about generalizing D315, but I
wanted a pattern for creating eyes in place (so, D7xx type) and not for
expanding eyespace (like do the D3xx ones).

Breakage : 7 PASSes, no FAILs

---- File [owl.tst]
262 PASSED                     possibly bogus
---- File [owl_rot.tst]
245 PASSED                     overlaps with Paul's patch
---- File [global.tst]
14 PASSED
---- File [13x13.tst]
149 PASSED
---- File [owl1.tst]
269 PASSED                     (the one I was targetting at first)
---- File [nngs2.tst]
150 PASSED
---- File [strategy5.tst]
232 PASSED                     possibly bogus

["possibly bogus" only means that I have no clue how the pattern influenced
  that result. It could be ok though]

Tuning pattern databases is very new to me. I'm not sure about the value for
instance, and it's quite possible that it could be generalized more. Could
it be more effective with constraints ? I'd be thankful for any hints and/or
comments.

/nando




reply via email to

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