[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[gnugo-devel] tuning patch
From: |
Gunnar Farneback |
Subject: |
[gnugo-devel] tuning patch |
Date: |
Sun, 02 Feb 2003 18:48:35 +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) |
A little bit of tuning. Regression delta:
trevorc:130 FAIL H6 [H9] Accidental
nngs3:630 PASS A17 [A17] Possibly properly solved
gunnar:15 PASS O6 [O6] Properly solved
nando:112 PASS O6 [!P7] Accidental
13x13b:26 PASS D3 [D3] Properly solved
13x13b:41 PASS J6 [J6] Properly solved
13x13b:42 PASS L13 [L13] Properly solved
The failure in trevorc:130 is annoying and not really the fault of
this patch. It's a horizon problem caused by the reduced depth reading
in do_compute_influence(), combined with some caching.
The pass in nando:112 is definitely accidental. We may wish to revise
that test case.
The patch also revises the incorrect test case connection:101.
- tuning
/Gunnar
Index: patterns/aa_attackpats.db
===================================================================
RCS file: /cvsroot/gnugo/gnugo/patterns/aa_attackpats.db,v
retrieving revision 1.11
diff -u -r1.11 aa_attackpats.db
--- patterns/aa_attackpats.db 15 Jan 2003 08:24:02 -0000 1.11
+++ patterns/aa_attackpats.db 2 Feb 2003 12:11:23 -0000
@@ -293,4 +293,24 @@
;lib(A)==3 && olib(*)>=2
+Pattern A18
+# gf New pattern. (3.3.17)
+
+??.?
+?X*.
+OOX.
+?X..
+??.?
+
+:8,-
+
+??.?
+?C*.
+eeXa
+?Db.
+??.?
+
+;!oplay_attack(*,e) && oplay_attack(*,a,C) && oplay_attack(b,a,D)
+
+
# END OF FILE
Index: patterns/barriers.db
===================================================================
RCS file: /cvsroot/gnugo/gnugo/patterns/barriers.db,v
retrieving revision 1.47
diff -u -r1.47 barriers.db
--- patterns/barriers.db 15 Jan 2003 08:24:02 -0000 1.47
+++ patterns/barriers.db 2 Feb 2003 12:11:25 -0000
@@ -366,15 +366,15 @@
Pattern Barrier23
-O,..
-..,O
+O,,.
+.,,O
:O,D
Oab.
.dcO
-;oplay_attack_either(a,b,c,d,b,d) || !oplay_defend_both(a,b,c,d,a,c)
+;oplay_attack_either(a,b,c,d,b,d) || oplay_defend_both(a,b,c,d,a,c)
Pattern Barrier24
@@ -2084,16 +2084,17 @@
Pattern Nonterritory1b
# gf Revised constraint. (3.3.10)
+# gf Revised constraint. (3.3.17)
Ox.
XoX
:8,t
-Ox.
+Oxb
XaX
->if ((halfeye(a) || false_eye(a))
+>if ((halfeye(a) || false_eye(a)) && safe_omove(b)
> && !false_eye_territory(a)) non_xterritory(a);
@@ -2729,6 +2733,11 @@
Pattern Nonterritory32
# gf New pattern. (3.1.23)
+# gf This pattern is somewhat problematic. O can often block at d to
+# secure e or block at b to secure c. The real point of the pattern
+# is that O can't secure both, but there's no way to express this
+# accurately. See gunnar:15 for a position where this pattern is
+# bad. (3.3.17)
..O
...
@@ -2997,6 +3006,30 @@
;lib(C)==2 && safe_omove(a) && oplay_attack(a,C) && !oplay_attack(a,b,a)
>non_xterritory(b);
+
+
+Pattern Nonterritory46
+# gf New pattern. (3.3.17)
+
+??X?
+O...
+X...
+?.X?
+
+:8,t
+
+??X?
+Oab.
+Xcef
+?dX?
+
+;oplay_defend(a,b,c,d,e,f,a)
+
+>non_xterritory(b);
+>non_xterritory(c);
+>non_xterritory(d);
+>non_xterritory(e);
+>non_xterritory(f);
# END OF FILE
Index: patterns/eyes.db
===================================================================
RCS file: /cvsroot/gnugo/gnugo/patterns/eyes.db,v
retrieving revision 1.32
diff -u -r1.32 eyes.db
--- patterns/eyes.db 2 Jan 2003 19:41:19 -0000 1.32
+++ patterns/eyes.db 2 Feb 2003 12:11:27 -0000
@@ -3699,6 +3699,14 @@
:0022
+Pattern 6551
+
+!.x
+!.x
+
+:0111
+
+
# ...
# Topology: ...
#
Index: patterns/fuseki.db
===================================================================
RCS file: /cvsroot/gnugo/gnugo/patterns/fuseki.db,v
retrieving revision 1.26
diff -u -r1.26 fuseki.db
--- patterns/fuseki.db 12 Jan 2003 15:26:53 -0000 1.26
+++ patterns/fuseki.db 2 Feb 2003 12:11:29 -0000
@@ -1024,8 +1024,8 @@
Pattern F408
-#FIXME: Make sure we're not inside our moyo
-# see 13x13:12
+# gf Added constraint to avoid overconcentration. See 13x13:12,
+# 13x13b:42. (3.3.17)
....... extend
oO..*..
@@ -1035,6 +1035,14 @@
:8,edt
+.....a.
+oO..*..
+.......
+.......
+-------
+
+;!omoyo(a)
+
Pattern F409
@@ -1529,16 +1537,17 @@
Pattern F709
+# gf Revised and devalued. See 13x13b:26.
-...... urgent jump
-..*..?
-X....?
-..O..X
-......
-......
-------
+?...... urgent jump
+?..*..?
+xX....?
+x..O..X
+x......
+.......
+-------
-:8,FEJ
+:8,FEj
############################################
Index: patterns/patterns.db
===================================================================
RCS file: /cvsroot/gnugo/gnugo/patterns/patterns.db,v
retrieving revision 1.99
diff -u -r1.99 patterns.db
--- patterns/patterns.db 15 Jan 2003 08:24:02 -0000 1.99
+++ patterns/patterns.db 2 Feb 2003 12:11:38 -0000
@@ -10153,6 +10153,10 @@
Pattern ED66
# tm Modified (3.1.18) (see trevorb:610)
# FIXME: Is this just too generic?
+# gf This pattern is way too general, see e.g. 13x13b:8.
+# Furthermore the second half of the constraint doesn't really make
+# sense. Consider removal. (3.3.17)
+
?x? draw back for safety
O.X
@@ -10177,6 +10181,7 @@
# tm Modified (3.1.18) (see trevorb:610)
# FIXME: Is this just too generic?
# Identical to ED66, but with j attribute instead of J
+# gf See comments for ED66. This pattern is redundant anyway. (3.3.17)
?x? draw back for safety
O.X
Index: regression/connection.tst
===================================================================
RCS file: /cvsroot/gnugo/gnugo/regression/connection.tst,v
retrieving revision 1.52
diff -u -r1.52 connection.tst
--- regression/connection.tst 25 Jan 2003 14:12:01 -0000 1.52
+++ regression/connection.tst 2 Feb 2003 12:11:39 -0000
@@ -334,8 +334,8 @@
#? [(2|3) J15]*
loadsgf games/nngs/Temnik-gnugo-3.3.15-200301161937.sgf 232
-101 disconnect N2 L5
-#? [1 M3]*
+101 disconnect O3 L5
+#? [1 (M3|N3)]*
# Report number of nodes visited by the tactical reading
10000 get_reading_node_counter
[Prev in Thread] |
Current Thread |
[Next in Thread] |
- [gnugo-devel] tuning patch,
Gunnar Farneback <=