[Top][All Lists]

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

Re: [gnugo-devel] Scoring problem

From: Gunnar Farneback
Subject: Re: [gnugo-devel] Scoring problem
Date: Wed, 16 Jul 2003 23:36:35 +0200
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)

Dan wrote:
> The aftermath code computes the score of this game
> incorrectly.

As usual a problem where the owl reading acts weirdly, finding a bogus
defense for a dead group, but only after the *opponent* has
strengthened his position locally.

> Correct is W+53.5 points. This is the score found
> by --score last or --score end. The game record
> shows W+27.5 points which was obtained by calling
> the gtp function final_score, based on the aftermath
> code. If you load the file with --score aftermath you
> get yet another score, 29.5 points.

No idea about the difference between 27.5 and 29.5. When loaded from
the game record both final_score and --score aftermath say W+29.5.

The simplest solution for the owl problem is the appended patch,
causing all scoring methods to agree on W+53.5. Use it in 3.4 if you
want, or postpone it for 3.5.1. I haven't checked what it does to
regressions or node counts.


Index: patterns/owl_attackpats.db
RCS file: /cvsroot/gnugo/gnugo/patterns/owl_attackpats.db,v
retrieving revision 1.90
diff -u -r1.90 owl_attackpats.db
--- patterns/owl_attackpats.db  4 Jul 2003 11:50:28 -0000       1.90
+++ patterns/owl_attackpats.db  16 Jul 2003 21:34:03 -0000
@@ -412,7 +412,7 @@
 Pattern A204
 ??O??              second line block

reply via email to

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