[Top][All Lists]

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

[gnugo-devel] (less) vital attack moves

From: Portela Fernand
Subject: [gnugo-devel] (less) vital attack moves
Date: Mon, 9 Dec 2002 14:45:14 +0100


The rationale behind this one-liner patch is simple : when the defender
hasn't (yet) got enough space to create 2 eyes, it is more important to
limit expansion and/or prevent escape, than to try to falsen the already
present eye, if any. So, vital attacking moves should be generated only
if the defender has got the potential to make at least 2 eyes.

Regression breakage is light : 2 PASSes and 2 FAILs

trevora:220             FAILED

  Accidental. This is the persistent reading cache I was talking about a
  while ago. B6 gets overvalued because of this and till now, I haven't
  been able to track it down.

strategy5:223   PASSED

  Good. The kind of success I was hoping for. GG looses less time with
  ineffective moves and finds a kill.

century2002:240 FAILED

  Accidental. Persistent owl cache effect, if the test is taken out of the
  batch, GG still plays F18.

arend2:50               passed

  I haven't investigated this one very much, but it looks good.

Else, there's a global improvement of a bit more than 1% owl nodes.


- no vital attack moves when defender is already short of eyespace.

Index: engine/owl.c
RCS file: /cvsroot/gnugo/gnugo/engine/owl.c,v
retrieving revision 1.124
diff -u -r1.124 owl.c
--- engine/owl.c        6 Dec 2002 23:31:42 -0000       1.124
+++ engine/owl.c        9 Dec 2002 10:26:10 -0000
@@ -2300,7 +2300,7 @@
       matchpat(owl_shapes_callback, other,
               &owl_vital_apat_db, dummy_moves, owl->goal);
-    else
+    else if (max_eyes(probable_eyes) >= 2)
       matchpat(owl_shapes_callback, other,
               &owl_vital_apat_db, vital_moves, owl->goal);

reply via email to

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