[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[gnugo-devel] two new patterns, one new test
From: |
Paul Pogonyshev |
Subject: |
[gnugo-devel] two new patterns, one new test |
Date: |
Thu, 14 Nov 2002 22:01:58 +0200 |
this patch addresses the failure in ld_owl:168. it also adds another test,
ld_owl:201, which is to check that gnu go now understands why Q19 is not good
in this position.
the pattern in owl_defendpats.db makes gnu go believe that T19 works. maybe
T19 is not the best solution (to me they all seem to be about equal
"quality"), but since it does work, gnu go should better understand why. as
T19 is tried first by owl code, it is now selected as an answer (good,
problem solved).
D1424 (the new pattern) is much like D818b, but i'm not confident enough to
start changing existing patterns. if anyone can merge D1424 into D818b, it
will be good.
another pattern is in owl_vital_apats.db. it is needed to overcome the
failure to see why "two certain eyes" can still be attacked. solves the new
testcase.
i also had another look at VA50. it gives at least seki (as long as black has
no other eyes), but as far as i understand, owl currently has no machinery to
deal with sekis. it can also kill black if he can be forced to fill one of
the two remaining liberties. however, we don't have machinery to determine
this either. my eyespace/connection code would do here, if it worked properly
;)
so i just wanted to ask, maybe it is better to comment out this pattern for
now, or tweak its valuation?
Paul
p.s. i didn't run the regression suit, but the patch can hardly break
anything.
Index: patterns/owl_vital_apats.db
===================================================================
RCS file: /cvsroot/gnugo/gnugo/patterns/owl_vital_apats.db,v
retrieving revision 1.31
diff -u -p -r1.31 owl_vital_apats.db
--- patterns/owl_vital_apats.db 20 Oct 2002 10:43:25 -0000 1.31
+++ patterns/owl_vital_apats.db 14 Nov 2002 19:57:26 -0000
@@ -842,4 +842,17 @@ Pattern VA50
:8,s,value(45)
+Pattern VA51
+# pp New pattern (3.3.12)
+
+|*Oo start an unfavourable ko
+|.XO
+|X.X
+|.Xx
+|.O.
++---
+
+:8,-,value(35)
+
+
# END OF FILE
Index: patterns/owl_defendpats.db
===================================================================
RCS file: /cvsroot/gnugo/gnugo/patterns/owl_defendpats.db,v
retrieving revision 1.72
diff -u -p -r1.72 owl_defendpats.db
--- patterns/owl_defendpats.db 11 Nov 2002 23:39:57 -0000 1.72
+++ patterns/owl_defendpats.db 14 Nov 2002 19:57:31 -0000
@@ -6599,4 +6599,15 @@ Pattern D1423
:8,s,value(45)
+Pattern D1424
+# pp New pattern (3.3.12)
+
+|Oo? avoid corner ko
+|.Oo
+|oX*
++---
+
+:8,-,value(45)
+
+
# END OF FILE
Index: regression/ld_owl.tst
===================================================================
RCS file: /cvsroot/gnugo/gnugo/regression/ld_owl.tst,v
retrieving revision 1.18
diff -u -p -r1.18 ld_owl.tst
--- regression/ld_owl.tst 30 Oct 2002 00:10:16 -0000 1.18
+++ regression/ld_owl.tst 14 Nov 2002 19:57:32 -0000
@@ -314,6 +314,14 @@ loadsgf games/life_and_death/tripod9.sgf
200 dragon_status C3
#? [dead]
+# More life and death problems around tripod group.
+
+loadsgf games/life_and_death/tripod4.sgf
+# Black tries a bad defense. He must be punished.
+play black Q19
+201 owl_attack Q19
+#? [3 T15]
+
# Mixed life and death problems.
loadsgf games/life_and_death/mixed1.sgf
300 owl_attack C17
[Prev in Thread] |
Current Thread |
[Next in Thread] |
- [gnugo-devel] two new patterns, one new test,
Paul Pogonyshev <=