Where to start?

[Rolf Neumann <rneumann@xxxxxxxxxxxxx>]

Hello together,

I am in research mode for a computer go project. It's not neccessary to
create the world best KI, but it should be good enough to satisfy
completely unexperienced and less experienced users.

I think the best strategy is to use some combination of brute force,
pattern matching and other more conventional solutions. I am not
interested in using neuronal networks and generic algorithms, but what I
want to do is to save all played games in a tree to give the program the
ability to remember his mistakes. I used this "brain"-method to improve
the weakness of the algorithm in some other projects. So the program can
learn a little bit and the user has the impression of KI.

I am not sure, whether it will work or not, because of the very high
number of possible variations. But I think it's possible to reduce them
a lot

1. not all moves make sense 
2. it's possible to turn and rotate the board

So the there are only 10 starting points on a 9x9 board, 21 on a 13x13
board and 45 on a 19x19 board. I am right? I don't think, it's very wise
to start on the border of the field.

The next thing, I thought about is not to look at the game in single
moves, stones and fields, but in tasks like fight, attack, defense,
expansion and so on. I "only" have to find out what the user intends
with his actual turn, examine the involved groups and decide whether and
how to react. Because lots of situations will appear more then once,
perhaps it's possible to give every situation an own "sub-brain".

What do you thing about that? Should I better forget it, or do you think
it's worth to give it a try?

Rolf Neumann