易野 <[EMAIL PROTECTED]> wrote: > In attached game record, GNUGO take white. After Black's move #17, #31, > White failed reply the position of #35 and #37.
So the problem is that when black plays R7 here, . . . . + . . . 10 . . . . . . . . 9 . . . . . . . . 8 . . . . . X . . 7 . . . . O O . . 6 . . . . . X . . 5 . . . . X . . . 4 . . . . . . . . 3 . . . . . . . . 2 . . . . . . . . 1 M N O P Q R S T GNU Go doesn't appreciate the value of separating black at S6. > That maybe come from an incorrect pathern match. Rather there's a missing pattern. Below is a patch adding a testcase for this problem and a pattern solving it. Someone stronger may suggest a more precise test. This problem is also documented at http://trac.gnugo.org/gnugo/ticket/69. /Gunnar Index: patterns/patterns.db =================================================================== RCS file: /cvsroot/gnugo/gnugo/patterns/patterns.db,v retrieving revision 1.139 diff -u -r1.139 patterns.db --- patterns/patterns.db 12 Jun 2005 09:34:15 -0000 1.139 +++ patterns/patterns.db 22 Jan 2006 22:46:54 -0000 @@ -11001,6 +11001,20 @@ ; weak(A) +Pattern ED100 +# gf New pattern. (3.7.8) + +..... cut through to weaken +X.O.. +.XOX. +..*.. +..... +----- + +:8,OXae + + + ##################################################################### # # Center defend/attack patterns Index: regression/strategy5.tst =================================================================== RCS file: /cvsroot/gnugo/gnugo/regression/strategy5.tst,v retrieving revision 1.49 diff -u -r1.49 strategy5.tst --- regression/strategy5.tst 31 Oct 2005 23:40:32 -0000 1.49 +++ regression/strategy5.tst 22 Jan 2006 22:46:54 -0000 @@ -175,6 +175,10 @@ 293 reg_genmove white #? [F17]* +# Very important to separate black. +loadsgf games/tso_igobot_2006-01-16.sgf 18 +294 restricted_genmove white O6 S6 +#? [S6] ############ End of Tests #################
_______________________________________________ gnugo-devel mailing list gnugo-devel@gnu.org http://lists.gnu.org/mailman/listinfo/gnugo-devel
