Re: [gnugo-devel] Joseki variations question

[Evan Berggren Daniel]

On Fri, 28 Mar 2003, Arend Bayer wrote:

> Objectively, Q2 is the only move here. For GNU Go, I'd also be somewhat
> satisfied iwth Q6.
>
> I think this position is too rare to make this worth a variation in the
> joseki database. Instead, we should teach GNU Go to understand that the
> lower right is extremely urgent. There are many similar but not
> identical positions. It's certainly worth a regression test.

OK, here's a test.  I checked, and locally gnugo knows that Q2 is the best
move.  restricted_genmove white Q6 P6 Q2 R2 gets things right.

I also included a test at move 11, where Gnu Go realises that a local move
is needed, but plays S4 instead of Q2 (which is still the move, right?).
Given that Gnu Go answers differently in the two cases, I think two tests
are in order.

Evan Daniel

Index: regression/nngs4.tst
===================================================================
RCS file: /cvsroot/gnugo/gnugo/regression/nngs4.tst,v
retrieving revision 1.23
diff -u -d -r1.23 nngs4.tst
--- regression/nngs4.tst        23 Mar 2003 13:30:56 -0000      1.23
+++ regression/nngs4.tst        27 Mar 2003 23:42:32 -0000
@@ -214,6 +214,13 @@
 520 restricted_genmove black J8 J9 H14 F13 E13 B16 G13 J15 C14
 #? [!J8|J9]

+loadsgf games/nngs/gnugo-3.3.17-ylamaki-200303262305.sgf 9
+530 gg_genmove white
+#? [Q2]
+
+loadsgf games/nngs/gnugo-3.3.17-ylamaki-200303262305.sgf 11
+540 gg_genmove white
+#? [Q2]


 ############ End of Tests #################