[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[gnugo-devel] arend_3_3.1: mixed patch
From: |
Arend Bayer |
Subject: |
[gnugo-devel] arend_3_3.1: mixed patch |
Date: |
Sat, 11 May 2002 01:49:03 +0200 (CEST) |
- copy-and-paste typo fixed in simple_ladder_attack
- influence and pattern tuning
- refutation of hoshi non-joseki added
- more tests
This is a collection of several unrelated matters, some of which had been
accumulating in the pre-3.2-release period.
IIRC Gunnar agreed that the call to find_defense in simple_ladder_attack is
a typo, and it does reduce the number of nodes in a particular example from
5.7*10^5 by about 40% (that was for a single ladder! -- the position was
admittedly very strange).
The tuning is very minor. The added influence pattern solves one of the new
test cases and breaks another existing test; overall I think it's more
dangerous not to have it than to have it.
The non-joseki is the following, where I added a few variations along the
following line (this might not be the most efficient punishment of the Black
mistake, but it looks good enough and solid for GNU GO).
X moves first:
+--------+
+........|
+........|
+..51OO2.|
+...4XX3.|
+....6...|
+.....O..|
+........|
+
The tests are: one move from the GNU Go vs GNU Go game Marco Scheurer
e-mailed to the list, the other some disastrous moves in a game I watched
GNU Go playing at WING. (The copyright notice at WING is identical to the
one of NNGS and gives permission to reproduce.)
Arend
Index: engine/reading.c
===================================================================
RCS file: /cvsroot/gnugo/gnugo/engine/reading.c,v
retrieving revision 1.60
diff -u -r1.60 reading.c
--- engine/reading.c 28 Apr 2002 14:46:43 -0000 1.60
+++ engine/reading.c 10 May 2002 23:16:03 -0000
@@ -6591,7 +6591,7 @@
}
}
else {
- if (do_find_defense(str, NULL, new_komaster, new_kom_pos) != WIN) {
+ if (simple_ladder_defend(str, NULL, new_komaster, new_kom_pos) != WIN) {
savemove = apos;
savecode = KO_B;
}
Index: patterns/barriers.db
===================================================================
RCS file: /cvsroot/gnugo/gnugo/patterns/barriers.db,v
retrieving revision 1.31
diff -u -r1.31 barriers.db
--- patterns/barriers.db 27 Apr 2002 22:08:58 -0000 1.31
+++ patterns/barriers.db 10 May 2002 23:16:07 -0000
@@ -1808,6 +1808,22 @@
;!oplay_attack_either(a,b,c,c,d)
+Pattern Intrusion 51
+
+?Q?
+X.X
+.!.
+
+:|,B,value(30)
+
+?Q?
+cad
+.b.
+
+;oplay_break_through(a,b,c,b,d)
+>return(!xplay_attack(a,b));
+
+
Pattern Nonterritory1
OxO
Index: patterns/hoshi.sgf
===================================================================
RCS file: /cvsroot/gnugo/gnugo/patterns/hoshi.sgf,v
retrieving revision 1.9
diff -u -r1.9 hoshi.sgf
--- patterns/hoshi.sgf 30 Mar 2002 23:44:44 -0000 1.9
+++ patterns/hoshi.sgf 10 May 2002 23:16:09 -0000
@@ -24,9 +24,9 @@
GN[Hoshi joseki variations]
DT[1999-05-10]
SY[Cgoban 1.9.5]TM[-];B[pd]C[0]
-(;B[pj];C[#kosumi tsuke for 6-9 stone games];W[qf]PL[B]
+(;B[pj];C[#kosumi tsuke for 6-9 stone games];W[qf]PL[1]
(;B[qe]MA[mk];W[pf]C[S]MA[mi]
-(;B[nd]MA[mk]C[U]PL[W];W[ph]C[j];B[re]MA[lk]C[j])
+(;B[nd]MA[mk]C[U]PL[2];W[ph]C[j];B[re]MA[lk]C[j])
(;B[tt];W[od]MA[mi];B[oc];W[nd]TR[mi];B[nc];W[pc]MA[mi];B[qc];W[pb]
MA[mi];B[qb];W[pe]MA[mi];B[qd])
@@ -80,8 +80,8 @@
(;W[qe]C[S]MA[mj];B[og]C[S]MA[mj])
)
-(;B[qd]C[S]MA[ij];C[#b blocks below];W[pc]C[S]MA[mj];B[od]C[S]MA[mi];
-W[rd]C[S]MA[mi];B[re]C[S]MA[mi];W[rc]C[S]MA[mi]
+(;B[qd]C[S]MA[ij];C[#b blocks below];W[pc]C[S]MA[mj]
+(;B[od]C[S]MA[mi];W[rd]C[S]MA[mi];B[re]C[S]MA[mi];W[rc]C[S]MA[mi]
(;B[qe]C[S]MA[mi];W[nc]C[S]MA[mi];B[pf]C[S]MA[mj])
(;B[oc]C[0
@@ -92,6 +92,22 @@
]MA[li])
)
+(;B[oc]C[# Non-joseki often played against GNU Go
+];W[rc]MA[lg]C[U
+];
+B[rd];W[od]MA[lg]C[U
+]
+(;B[nc];W[pe]MA[lg]C[S
+];B[oe];W[nd]MA[lg]C[U
+])
+
+(;B[oe];W[nd]MA[lg]C[U
+];B[nc];W[pe]MA[lg]C[U
+];B[of];W[md]MA[kg]C[U
+])
+)
+)
+
(;B[pf]MA[nj]C[0
])
)
@@ -200,26 +216,26 @@
(;W[qc];B[qd]MA[mg])
)
-(;B[pj]C[0]PL[B];B[pf]MA[ml];C[#tsuke nobi 6-9 stone games];W[pg]
+(;B[pj]C[0]PL[1];B[pf]MA[ml];C[#tsuke nobi 6-9 stone games];W[pg]
MA[mi]C[U];B[of]MA[ml]C[U];C[#tsuke nobi 6-9 stone games])
(;B[pf]MA[mi]C[j
];C[#tsuke nobi normal]
(;W[pg]MA[mi]C[U]
(;B[of]MA[mh]C[U];C[#nobi]
-(;W[qe]MA[mi]C[U];C[#w crawls];B[qd]MA[mh]C[U]PL[W]
+(;W[qe]MA[mi]C[U];C[#w crawls];B[qd]MA[mh]C[U]PL[2]
(;W[qj]MA[nk]C[U];C[#standard joseki];B[nd]MA[mh]C[U])
(;W[qh]MA[nk]C[S];C[#alternative joseki];B[nd]MA[mh]C[U
-]PL[B]
-(;W[pe]PL[W]
+]PL[1]
+(;W[pe]PL[2]
(;B[oe]MA[mh])
(;W[oe];B[od]MA[lh]C[S
])
)
-(;B[jd]PL[B];B[jc]MA[if]PL[B]
+(;B[jd]PL[1];B[jc]MA[if]PL[1]
C[t
#Usually good in high handicap games (HL 2-nov-2000)
];B[jf]C[t
@@ -409,7 +425,7 @@
]MA[jj];B[rb]
C[0];W[pb]MA[kg]C[S];B[qb]C[0];W[oc]C[S]MA[kg])
-(;W[tt]PL[W]
+(;W[tt]PL[2]
(;B[pj];W[ph]C[S]MA[mk])
(;W[qk]C[0];B[tt];W[ph]C[S]MA[ml])
Index: patterns/patterns.db
===================================================================
RCS file: /cvsroot/gnugo/gnugo/patterns/patterns.db,v
retrieving revision 1.72
diff -u -r1.72 patterns.db
--- patterns/patterns.db 10 May 2002 17:18:59 -0000 1.72
+++ patterns/patterns.db 10 May 2002 23:16:25 -0000
@@ -8823,6 +8823,7 @@
Pattern CB328
+#ab Revised constraint (3.3.3)
?.? seal opponent in and connect
...
@@ -8836,7 +8837,7 @@
a*O
OB?
-; oplay_attack(*,a,a) && (!attack(B) || does_attack(*,B))
+; oplay_attack_either(*,a,a,B) && (!attack(B) || does_attack(*,B))
#tm - removed this pattern (3.1.20)
Index: patterns/patterns2.db
===================================================================
RCS file: /cvsroot/gnugo/gnugo/patterns/patterns2.db,v
retrieving revision 1.37
diff -u -r1.37 patterns2.db
--- patterns/patterns2.db 3 May 2002 17:57:58 -0000 1.37
+++ patterns/patterns2.db 10 May 2002 23:16:30 -0000
@@ -1538,6 +1538,8 @@
Pattern Shape13
+# Such huge shape values can cause problems. See handtalk12.sgf, move 37
+# (where * is actually an owl defense). -Arend
?X? don't set up double cut
*.O
Index: regression/owl.tst
===================================================================
RCS file: /cvsroot/gnugo/gnugo/regression/owl.tst,v
retrieving revision 1.51
diff -u -r1.51 owl.tst
--- regression/owl.tst 6 May 2002 23:38:36 -0000 1.51
+++ regression/owl.tst 10 May 2002 23:16:32 -0000
@@ -951,6 +951,17 @@
262 owl_defend E7
#? [0]*
+# Extremely bad misread. (Made GNU Go play at L9 with the last move before
+# that position.)
+loadsgf games/wing-yuhiko-gnugo.sgf 115
+263 owl_defend M13
+#? [1 R11]
+
+# Need another vital attack pattern here?
+loadsgf games/wing-yuhiko-gnugo.sgf 220
+264 owl_attack D18
+#? [2 A18]
+
########### end of tests #####################
# Report number of nodes visited by the tactical reading
Index: regression/strategy4.tst
===================================================================
RCS file: /cvsroot/gnugo/gnugo/regression/strategy4.tst,v
retrieving revision 1.45
diff -u -r1.45 strategy4.tst
--- regression/strategy4.tst 6 May 2002 23:38:36 -0000 1.45
+++ regression/strategy4.tst 10 May 2002 23:16:34 -0000
@@ -352,6 +352,29 @@
220 gg_genmove white
#? [R6]*
+loadsgf games/wing-yuhiko-gnugo.sgf 12
+221 gg_genmove black
+#? [S4|C11]
+
+# This might be worth a J pattern.
+loadsgf games/wing-yuhiko-gnugo.sgf 30
+222 gg_genmove black
+#? [R14]
+
+loadsgf games/wing-yuhiko-gnugo.sgf 46
+223 gg_genmove black
+#? [P16]
+
+#CATEGORY=STRATEGY
+loadsgf games/wing-yuhiko-gnugo.sgf 62
+224 gg_genmove black
+#? [N15]
+
+#CATEGORY=STRATEGY
+loadsgf games/wing-yuhiko-gnugo.sgf O12
+225 gg_genmove black
+#? [N15]
+
############ End of Tests #################
- [gnugo-devel] arend_3_3.1: mixed patch,
Arend Bayer <=