gnugo-devel
[Top][All Lists]
Advanced

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

[gnugo-devel] Possible test cases


From: Tom Schumm
Subject: [gnugo-devel] Possible test cases
Date: Fri, 23 Apr 2004 17:15:47 -0500
User-agent: KMail/1.6.1

I've got a nightly CVS compile of gnugo running on KGS (username GnuGoCVS) and
a couple times while watching it I've noticed some particularly dubious play
that might make useful test cases:

The first (2004-04-10-Shun9137-GnuGoCVS.sgf) is a useless throw-in in a dead
group.  I was talking with Evan and he suggested it was probably a
thrashing_dragon situation.  Even if the group had a chance to live though,
the move made would not have been a useful attack on it and wasn't sente.
Don't know if it's worth adding as a test case or not...

The second case (2004-04-10-GnuGoCVS-inuyasha34.sgf) is a more interesting...
The engine misplayed a ladder which happened to intersect with another ladder
in progress.  It was triple-naughty for allowing the ladder to occur in the
first place, misplaying it, and not cleaning up the other ladder earlier at
several opportunities.  Using -t the engine even detects that running in the
ladder is a blunder but still has a really high value for the move.

Also, early in the same game there were some really terrible moves and tenukis
in a fight in the upper right.

I've written up a few tests.  I don't know which (if any) of them are most
interesting to include.  I'll suggest calling it "kgs.tst" and putting the
games in games/kgs unless someone else has a better place for them.

Oh, and I probably ought to introduce myself since I'm new here.  You might
know me as fwiffo on kgs and nngs.  I intend to actually get into the code at
some point instead of just complaining.  :-)

-- 
Tom Schumm
http://www.phong.org/

Attachment: kgs.tst
Description: Text document

Attachment: 2004-04-10-GnuGoCVS-inuyasha34.sgf
Description: application/go-sgf

Attachment: 2004-04-10-Shun9137-GnuGoCVS.sgf
Description: application/go-sgf


reply via email to

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