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

computer-go: Strange behaviour in Joseki



I try to find the right handicap between me and Gnugo 3.2. I'm about 8k,
but can't be beaten by it even at 9 stones. One thing I can't understand
is the following :


5 |
  |
4 |       B
  |
3 |    W3 B4    W1    B2
  |
2 |
  |
1 ----------------------
 A  B  C  D  E  F  G  H

In this hoschi sequence, Gnugo invariably plays this silly B4. This is
clearly the wrong direction, for B2 is far too near the wall B intend to
build. And it doesn't work with the stone further up (around D10) that
justify the strong pincer.

So my questions are :

- First do you know how this joseki could be included in Gnugo database,
and how to remove it ?
- Second, I know computers don't understand anything about go, but why
does it, from all those nice sequences, allways choose to pincer and then
fail in the subsequent joseki ? How do the program choose between
different choices ? Can those choices be enhanced ?
- Third, how does the joseki database work ? Is it only a sequence, or is
there additionnal datas for the program to perform its choice ?

I know this would be better in Gnugo dev' list. Please don't be specific
to gnugo, any academic answer will fit.

\bye

-- 

Nicolas FRANCOIS
http://nicolas.francois.free.fr

We are the Micro$oft.
Resistance is futile.
You will be assimilated.