RE: [Bug-gnubg] race-, position-database etc.

[Albert Silver]

> >Also, I find it hard to belive the bearoff database is insuffcient.
> >Can you pose this Q to Neil? i.e. can he provide a set of positions
GNU
> >mishandles?
> 
> I think there's a bit of a misunderstanding here. As far as I recall
this
> discussion on GOL, Neil Kazaross was thinking of SNOWIE 3.2 here (Neil
has
> only just started to use GNUBG seriously). SW seems to play races and
> bear-ins poorly, and perhaps even bear-offs (cube actions?).

I have seen mistakes. I have been using Dueller to play matches between
the two programs, and have had both programs analzye the matches. Snowie
has highlifghted a few mistakes which were confirmed by both programs
subsequently. I'd have to find them now, but I can definitely comfirm
that GNU has made mistakes, however few. 

Just to reclarify, this wasn't the fruit of a discussion at GoL, but
from a talk I had with him at GG.

                                                Albert

> 
> However, my experience with GNUBG and its RACE net and databases is
that
> together these are very accurate already (and fast), even at 0-ply!
> 
> I'm pretty sure there's very little room for improvement in this area
with
> GNUBG. Seems to me it'd be much more efficient to try and improve
GNUBG in
> other areas (like the training of the CRASHED net done right now).
> 
> 
> --
> Robert-Jan (Zorba on FIBS)
> 
> 
> 
> _______________________________________________
> Bug-gnubg mailing list
> address@hidden
> http://mail.gnu.org/mailman/listinfo/bug-gnubg