gnugo-devel
[Top][All Lists]
Advanced

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

[gnugo-devel] various tuning


From: Arend Bayer
Subject: [gnugo-devel] various tuning
Date: Tue, 8 Oct 2002 19:37:27 +0200 (CEST)

This patch contains mixed tuning. I tuned against 3.3.9, and the breakage
(here I've removed those casses that also PASS or FAIL in current CVS, but
did not check for new unexpected results) is 15 PASSes and 7 FAILs.

I have now disabled CB107 (the replacement pattern that caused the NNGS
crash), but I cannot say I am already completely satisfied with GNU Go's
behaviour in nngs1:21. The pattern that helps here is

+Pattern Intrusion54
+# ab New pattern (3.3.10)
+# X cannot play double hane.
+
+..X?
+.XQo
+!.oo
+
+:8,B,value(40)
+
+.fH?
+gXQb
+edac
+
+;(o_somewhere(a) || o_somewhere(b) || o_somewhere(c))
+;&& (o_somewhere(a) || oplay_attack(d,a,a)) && oplay_attack(f,g,H)

This makes a move below "!" less attractive.

But M6 in nngs1:21 is still the second-best move, and it is only because
of an in my opinion overvalued 't' pattern at G3 that GNU Go does not play
at M6. Adding a bad shape pattern for such keimas, as Dan suggested, would
stil make sense.

ALL the FAILs (but also a big number of PASSes) in the regression delta
are caused by my modification to A1124 (I made it more specific). However,
in all the positions where it came up, it seemed non-sense to play, so
I think the change is right and just the noise is high
(a typcial situation would be

XO.
O.*
.OX

where A1124 would make O to play at * to prevent X escaping to his friend
at the upper left).

In some cases this actually meant that there was no attacking move generated
at all, which sometimes means a change to the better and sometimes to the
worse. But then we should find better patterns for these situations.

Arend


./regress.sh . owl.tst
23 unexpected PASS!
180 unexpected PASS!
262 unexpected FAIL: Correct '0', got '1 E8'
./regress.sh . nngs1.tst
8 unexpected PASS!
35 unexpected PASS!
./regress.sh . neurogo.tst
12 unexpected FAIL: Correct 'B10', got 'S8'
./regress.sh . viking.tst
10 unexpected PASS!
./regress.sh . nicklas1.tst
1213 unexpected FAIL: Correct 'N4', got 'F12'
./regress.sh . nngs.tst
300 unexpected PASS!
./regress.sh . 13x13.tst
15 unexpected FAIL: Correct 'C7|B7|C6|B6|B5|C5', got 'E8'
16 unexpected PASS!
21 unexpected PASS!
149 unexpected FAIL: Correct 'L7', got 'L9'
./regress.sh . owl1.tst
283 unexpected PASS!
./regress.sh . nngs3.tst
670 unexpected PASS!
680 unexpected PASS!
710 unexpected FAIL: Correct 'Q2', got 'C2'
750 unexpected FAIL: Correct 'L17|L18', got 'K18'
./regress.sh . century2002.tst
200 unexpected PASS!
261 unexpected PASS!
./regress.sh . safety.tst
2 unexpected PASS!
5 unexpected PASS!


Index: patterns/barriers.db
===================================================================
RCS file: /cvsroot/gnugo/gnugo/patterns/barriers.db,v
retrieving revision 1.40
diff -u -r1.40 barriers.db
--- patterns/barriers.db        4 Oct 2002 09:14:12 -0000       1.40
+++ patterns/barriers.db        8 Oct 2002 17:05:01 -0000
@@ -547,6 +547,14 @@

 :8,A

+OAbx
+.cdx
+..,x
+----
+
+;lib(A)>3 || (!legal_omove(b))
+;|| (lib(A)>2 && (olib(b) <= 2 || oplay_attack_either(c,d,b,b,c)))
+

 Pattern Barrier38

@@ -927,6 +935,25 @@
 ;!oplay_attack(a,b,c,c)


+Pattern Barrier65
+# ab New (3.3.10)
+
+?O.
+X,.
+x,.
+---
+
+:8,OD
+
+?O.
+Ca.
+xb.
+---
+
+;oplay_attack(a,b,C)
+
+
+
 Pattern Intrusion1

 xxxx
@@ -1435,6 +1462,12 @@

 :8,B,value(30)

+ba.
+Q!!
+---
+
+>return (!xplay_attack(a,b));
+

 Pattern Intrusion25

@@ -1943,6 +1976,24 @@
 QB!

 ;!oplay_defend(a,B)
+
+
+Pattern Intrusion54
+# ab New pattern (3.3.10)
+# X cannot play double hane.
+
+..X?
+.XQo
+!.oo
+
+:8,B,value(40)
+
+.fH?
+gXQb
+edac
+
+;(o_somewhere(a) || o_somewhere(b) || o_somewhere(c))
+;&& (o_somewhere(a) || oplay_attack(d,a,a)) && oplay_attack(f,g,H)


 Pattern Nonterritory1
Index: patterns/owl_attackpats.db
===================================================================
RCS file: /cvsroot/gnugo/gnugo/patterns/owl_attackpats.db,v
retrieving revision 1.60
diff -u -r1.60 owl_attackpats.db
--- patterns/owl_attackpats.db  29 Sep 2002 13:48:22 -0000      1.60
+++ patterns/owl_attackpats.db  8 Oct 2002 17:05:08 -0000
@@ -1487,6 +1487,23 @@

 ;owl_escape_value(a)>0 || owl_escape_value(b)>0

+Pattern A422
+# ab New pattern. (3.3.10)
+
+?Oo?       prevent connection to another dragon
+X*.X
+?xo?
+
+:8,-,value(85)
+
+?Oo?
+a*.b
+?xo?
+
+;(!owl_goal_dragon(a) || !owl_goal_dragon(b))
+;&& !xplay_disconnect(*,a,b)
+
+

 Pattern A422
 # db New pattern (3.3.9)
@@ -3504,6 +3521,7 @@
 Pattern A1124
 # tm New Pattern (3.1.22) (see nngs:770 decide-dragon K4)
 #   see also D1140
+# ab Revised constraint. (3.3.10)

 ?OX
 ?.*
@@ -3516,8 +3534,7 @@
 cd?

 ;(owl_escape_value(d) > 0
-; || owl_escape_value(b) > 0
-; || owl_escape_value(c) > 0)
+; || owl_escape_value(b) > 0)
 ;&& oplay_defend_both(*,e,A,*)


Index: patterns/owl_defendpats.db
===================================================================
RCS file: /cvsroot/gnugo/gnugo/patterns/owl_defendpats.db,v
retrieving revision 1.68
diff -u -r1.68 owl_defendpats.db
--- patterns/owl_defendpats.db  24 Sep 2002 19:11:54 -0000      1.68
+++ patterns/owl_defendpats.db  8 Oct 2002 17:05:17 -0000
@@ -4218,6 +4218,7 @@


 Pattern D1121
+# FIXME: If there is a X stone at "?", then this pattern is nonsense.

 OoXo       push into the open
 ?*.o
Index: patterns/patterns.db
===================================================================
RCS file: /cvsroot/gnugo/gnugo/patterns/patterns.db,v
retrieving revision 1.84
diff -u -r1.84 patterns.db
--- patterns/patterns.db        27 Sep 2002 00:47:17 -0000      1.84
+++ patterns/patterns.db        8 Oct 2002 17:05:34 -0000
@@ -825,7 +825,7 @@
 ;odefend_against(*,a)


-Pattern CC52
+Pattern CC52a
 # Constraint probably needs revision.

 OX..     connect if invaded
@@ -840,6 +840,24 @@
 ;&& oplay_attack(*,a,b,e,e) && oplay_attack(*,a,b,c,c)


+Pattern CC52b
+# ab New pattern. (3.3.10)
+
+?????
+.OX.?     keep loose connection
+..*.O
+?..oo
+
+:8,c
+
+?????
+.aX.?
+..*.b
+?..oo
+
+;!same_dragon(a,b) && (alive(a) || alive(b))
+
+
 Pattern CC53

 XXO           cut
@@ -1184,6 +1202,7 @@


 Pattern CC79
+# ab modified (3.3.10)

 ?O.xx
 XO.*.
@@ -1197,7 +1216,7 @@
 XXO..
 .....

-; lib(a)>=3 && !oarea(*)
+; lib(a)>=3 && !omoyo(*)


 Pattern CC80
@@ -7098,24 +7117,25 @@
 :|,OE,shape(2),jump_out_far_helper


-Pattern CB107
-# db New pattern (3.1.28)
-# db corrected (3.1.28)
-
-?OXxxx       defend the cut before leaping
-..O...
-?*....
-?.....
-
-:8,O
-
-?OAxxx
-.bO...
-?*c.a.
-?.....
-
-; oplay_defend(a,b,c,b) && !attack(A)
-> replace(a,*)
+#Pattern CB107
+## db New pattern (3.1.28)
+## db corrected (3.1.28)
+## ab Removed (3.3.10). (Now less important due to Intrusion54.)
+#
+#?OXxxx       defend the cut before leaping
+#..O...
+#?*....
+#?.....
+#
+#:8,O
+#
+#?OAxxx
+#.bO...
+#?*c.a.
+#?.....
+#
+#; oplay_defend(a,b,c,b) && !attack(A)
+#> replace(a,*)


 ###############





reply via email to

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