Re: [gnugo-devel] fuseki db

[Douglas Ridgway]

Hi Gunnar,

Thanks for taking a look at this, and sorry about the slow response.

On Sun, 21 Nov 2004, Gunnar Farnebäck wrote:

> Doug wrote:
> > In addition to the new .db files, I have included patches here to
> > raise some limits, remove the fuseki.c popularity screen and reduce
> > the pattern size to use less memory.  Still, the compiled binary is
> > larger: 7.5M, as as compared to 4.4M for 3.6-pre4.  If there are
> > situations where binary size is critical, a trimmed version could be
> > generated.
> 
> We might consider more drastic changes as well, like using 64-bit
> Zobrist hashes to identify the patterns. 

I'm not sure I understand. You mean for matching the fuseki patterns
during play, right? In principle, one could do all pattern matching with
hashes, but I agree it makes particular sense for whole-board patterns.

> It could also be an idea to
> distribute the patterns in a compressed format rather than .db format.
> (It should of course be possible to unpack to .db format, either as
> part of the build or only for inspection.)

This did occur to me, but since there's normally some general 
purpose compression applied along the way, it wasn't clear how big the 
benefit might be.

> >     488 (actually dominates most alternatives based on amateur results)
> >     616 (does fine, is by far most popular)
> >     649 (ditto)
> >     714 (followup to 616. This move is better than ave, but see below)
> >     1000 (seems ordinary)
> 
> Do you know if the joseki(?) in 616, 649, 714 appears in professional
> games, although in other full-board contexts?

I did a Gobase search on the 9x9 corner. The attach-and-extend joseki and
attach-and-block joseki are the most popular responses to the hane, of
course, but Gobase has five games where the hane is answered with another
hane, as in 616 and 649. Two are old Chinese games which start with
crossed star points. Of the three modern games, it's been played by
Takeuchi Kazuo against Kitani in 1960 (lost), Chin Kaei against Ma
Xiaochun in 1981 (lost), and Cho HunHyun against Yang JaeHo (won).  
Following up with another hane, as in 714, occurs only in one of the old
Chinese games. So pros know the move exists, and do play it, though 
rarely.

One other thing I noticed: if pattern 1000 is put back in, regression test
nngs4:810 ought to be changed as well. I guess I should check regression
for the whole new fuseki db, in case there are others like that.

So should I make a patch putting those five patterns back in, or just edit
the db files?

doug.