[gnugo-devel] arend_1_28.1: regression tests

[Arend Bayer]

 - more GoFun regression tests
 
This is only three tests, but I think there are serious blunders. I hope
my "invasion penalty" patch (on which I am working at the moment) can take
care of no. 34; no. 36 is a very difficult owl problem.

But no. 35 is really worrying, and I have seen similar bad moves recently.
I consider this a genuine territory valuation problem, and is maybe typical
for the weaknesses of --experimental-influence in the fuseki.

It should be solved before 3.2. 

Arend


Index: regression/nngs1.tst
===================================================================
RCS file: /cvsroot/gnugo/gnugo/regression/nngs1.tst,v
retrieving revision 1.10
diff -u -r1.10 nngs1.tst
--- regression/nngs1.tst        21 Mar 2002 14:23:09 -0000      1.10
+++ regression/nngs1.tst        23 Mar 2002 23:37:49 -0000
@@ -133,3 +133,16 @@
 loadsgf games/nngs/gnugo-3.1.27-Yuuki-200203132324.sgf 90
 33 gg_genmove white
 #? [!S9]
+
+loadsgf games/nngs/gnugo-3.1.27-GoFuN-200203230525.sgf 30
+34 gg_genmove white
+#? [!N10|N9|N11|O11|O10]*
+
+loadsgf games/nngs/gnugo-3.1.27-GoFuN-200203230525.sgf 44
+35 gg_genmove white
+#? [D13]*
+
+loadsgf games/nngs/gnugo-3.1.27-GoFuN-200203230525.sgf 64
+36 gg_genmove white
+#? [N11|L7]*
+