Re: [gnugo-devel] fast_defense() improvement

[Gunnar Farnebäck]

Arend wrote:
> I think this is besides the point. I "proved" it under the simplifying
> assumption that there is one move to connection all strings in question
> at once. Of course, this is not always the case, but other cases should
> be rare, and even then I would make a high bet that GNU Go as the
> attacker would give up (due to liberty cutoffs) a couple of moves later.

There's at least one under-the-stones tesuji that this code gets
wrong, namely:

 4 . . . . . . . . . 4
 3 . . O O O X X X . 3
 2 . . O X X O O X . 2
 1 . . O . O O X X . 1
   A B C D E F G H J

Currently fast_defense() thinks that black's small string is saved
after the capture at D1. The patch below adds this as test case
reading:199.

- new test case

/Gunnar

Index: regression/reading.tst
===================================================================
RCS file: /cvsroot/gnugo/gnugo/regression/reading.tst,v
retrieving revision 1.70
diff -u -r1.70 reading.tst
--- regression/reading.tst      8 Jun 2004 13:27:23 -0000       1.70
+++ regression/reading.tst      17 Jun 2004 00:34:22 -0000
@@ -835,6 +835,10 @@
 popgo
 popgo
 
+loadsgf games/reading45.sgf
+199 defend D2
+#? [0]
+
 # Report number of nodes visited by the tactical reading
 10000 get_reading_node_counter
 #? [0]&
Index: regression/games/reading45.sgf
===================================================================
RCS file: regression/games/reading45.sgf
diff -N regression/games/reading45.sgf
--- /dev/null   1 Jan 1970 00:00:00 -0000
+++ regression/games/reading45.sgf      17 Jun 2004 00:34:22 -0000
@@ -0,0 +1,10 @@
+(;GM[1]FF[3]
+RU[Japanese]SZ[9]HA[0]KM[5.5]
+PW[White]
+PB[Black]
+GN[White (W) vs. Black (B)]
+DT[2004-06-16]
+SY[Cgoban 1.9.14]TM[5:00:00(5x1:00)];
+AW[cg][dg][eg][ch][fh][gh][ci][ei][fi]
+AB[fg][gg][hg][dh][eh][hh][gi][hi]
+)