[Top][All Lists]

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

Re: [gnugo-devel] Patch arend_1_14.2 -- Missing endgame pattern

From: Gunnar Farneback
Subject: Re: [gnugo-devel] Patch arend_1_14.2 -- Missing endgame pattern
Date: Mon, 12 Nov 2001 17:56:51 +0100
User-agent: EMH/1.14.1 SEMI/1.14.3 (Ushinoya) FLIM/1.14.2 (Yagi-Nishiguchi) APEL/10.3 Emacs/20.7 (sparc-sun-solaris2.7) (with unibyte mode)

Arend wrote:
> Here is a basic endgame pattern that I found to be missing. I also added
> the test where it is necessary. Should be very precise with the
> constraints.

Good timing. I was just about to add two test cases from a recent NNGS
game, one of which I think exactly shows the lack of this pattern. The
patch below adds the two test cases, which I've also added to CVS.

> New file: (Don't know how to do that with cvs)

Usually the solution is to use "cvs add" to add the file to the
repository, then do "cvs diff -uN" to include the new file in the
patch. However, I guess this doesn't work with read-only access. I
don't know if there's a workaround then. The solution you chose is
manageable in any case. With many new files the best solution is
probably to package them in a tar archive.


> > Pattern EB715
> > # Added for 3.1.14
> > 
> > O.*    
> > X..
> > ---
> > 
> > :8,OXb
> > 
> > ac*
> > Xbd
> > ---
> > 
> > ;oplay_attack_either(b,a,b) && oplay_attack_either(c,b,d,c,d)

The constraint looks good, but it may be worth commenting that there's
an alternative solution, which might be better. If we look at the
first half of the constraint, the effective move (to attack either
string) is usually at c. Then an alternative is to use the


These differ in two relevant aspects:

1. attack_either() is somewhat weak at coming up with effective moves.
   It may miss the cutting move at c. Then the defend_both()
   construction has better chances to analyze correctly.
2. If the effective move is something other than c, then the
   defend_both() construction will fail.

Of course the same argument holds for the second half of the


Index: regression/strategy4.tst
RCS file: /cvsroot/gnugo/gnugo/regression/strategy4.tst,v
retrieving revision 1.23
diff -u -r1.23 strategy4.tst
--- regression/strategy4.tst    2001/11/10 09:26:25     1.23
+++ regression/strategy4.tst    2001/11/12 16:24:24
@@ -310,6 +310,14 @@
 212 gg_genmove white
 #? [G2]
+loadsgf games/strategy46.sgf 172
+213 gg_genmove white
+#? [F1]
+loadsgf games/strategy46.sgf 174
+214 gg_genmove white
+#? [D2]
 ############ End of Tests #################
 # Report number of nodes visited by the tactical reading
Index: regression/games/strategy46.sgf
RCS file: strategy46.sgf
diff -N strategy46.sgf
--- /dev/null   Sat Apr 14 20:46:23 2001
+++ strategy46.sgf      Mon Nov 12 11:24:24 2001
@@ -0,0 +1,33 @@
+PW[GnuGo (GNU Go 3.1.13)]WR[14k*]
+C[final_score: B+21.5
+M11 removed
+P8 removed
+Q6 removed
+A2 removed])

reply via email to

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