[Top][All Lists]

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

Re: [gnugo-devel] Pattern updates

From: bump
Subject: Re: [gnugo-devel] Pattern updates
Date: Thu, 3 Jul 2003 16:13:56 -0700

> I'm not sure leaving EJ61 there is more safe. In the case you posted earlier
> (against 3.3.16), it happened to let a group dead. In practice more
> posibilities exist for the connection, especially when the group is in
> critical state. Without EJ61, I didn't found any regression breakage. If we
> keep EJ61, should I add another pattern like EJ61b to make another kind of
> connection?

I agree that removing EJ61 is the right thing to do if GNU
won't forget to connect. And if we weren't close to 3.4 I
would have no objection.

In the game against 3.3.16 the right thing is to connect, but
with the diagonal connection (tiger). So you are suggesting
adding a pattern for that.

How about the following. The patch below turns off the pattern
if the inside dragon is critical. In the position in the game
against 3.3.16 this has the effect of disabling the pattern and
the owl code finds the move at O18.

The rationale is that if the owl code is aware of an issue
then let the owl code decide how to connect. Otherwise,
connect automatically.

If you change the position a little like this:

GN[GNU Go 3.3.22 load and print]
KM[5.5]AP[GNU Go:3.3.22]RU[Japanese];PL[W]

then the pattern is matched and GNU connects automatically.

Index: patterns/patterns.db
RCS file: /cvsroot/gnugo/gnugo/patterns/patterns.db,v
retrieving revision 1.113
diff -u -r1.113 patterns.db
--- patterns/patterns.db        27 Jun 2003 15:37:17 -0000      1.113
+++ patterns/patterns.db        3 Jul 2003 23:10:52 -0000
@@ -12543,6 +12543,14 @@
+; !critical(A)

reply via email to

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