gnugo-devel
[Top][All Lists]
Advanced

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

Re: [gnugo-devel] Gifu test suite.


From: Evan Berggren Daniel
Subject: Re: [gnugo-devel] Gifu test suite.
Date: Mon, 25 Aug 2003 21:19:24 -0400 (EDT)

On Mon, 25 Aug 2003, SP LEE wrote:

>
>
> > Below follows a test suite with 41 tests from the Gifu games. Please
> > review and extend.
> >
> > /Gunnar
> >
> >
> > # P14 is an atari which only helps black. Q14 is the only move.
> > loadsgf games/cgf2003/GnuGo-Haruka.sgf 10
> > 501 reg_genmove white
> > #? [Q14]
> >
>
> In the game Haruka started a battle in very early stage. Is there a
> joseki for this kind of cut? If this is too difficult to implement,
> there are several strategies we could implement.
> 1. Jump instead of knight jump to avoid any cut.
> 2. Avoid the joseki's which produce many (unsafe) dragons.
> 3. Make decision to sacrifice or exchange in the early stage. Although
> this seems to be difficult, there is often no other choice during a
> battle.

First, this joseki should probably be in the database if gnugo is going to
play it.  I also rather doubt that gnugo should play it.  The interesting
thing is, gnugo didn't really think it was playing it -- it thought it
was playing a 5,3 joseki, because the constraint in mokuhazushi.sgf is too
narrow.  There are no joseki responses to the two space low pincer given
in keima.sgf; this should probably be fixed.

However, we also need a more general solution to the problem; similar
positions will occur in other than joseki contexts.

Also worth noting is that without the mokuhazushi joseki, gnugo would have
played tenuki; P15 ranks highly, but would be unlikely simply because a
lot of other corner moves do too.

Evan Daniel




reply via email to

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