Re: Another Domain?

[Matt Gokey <mgokey@xxxxxxxxxx>]

Eric Moyer wrote:

> Using "atomic moves" of a single color is the same as deciding on a goal.
> (You decide what shape you would eventually like to make.)  It would
> then require searching (albeit limited) to find tactics to enact your
>strategy.

>Using "atomic moves" which consist of two colors of stones is the same
>as using a huristic on the tree search.  It compresses the tree and
>implicitly prunes.  Depending on the quality of your "atomic moves" this
>pruning could be good or bad.  As I understand pattern matching, this
>seems very similar to the standard technique.  The standard technique
>says: if I see x pattern here, search more intently in some other
>predefined place.  Your method is: if I and enemy combine to make part
>of pattern x in the first several moves, look at the outcome of making
>the rest of pattern x.

If I understand you correctly you are talking about finding moves and
strategies to reach locally defined goals (shape or otherwise, ie. cut,
capture, run, etc...). 

What I am implying is to skip that process and jump right to the
attained goal using multiple distinct paths for each goal.  These would
be moves for the current position. This would be repeated for the new
positions: defining goals, creating assumed paths to reach them which
gets to the next deeper layer of positions and the process repeats
itself.  So the "search" would be done using assumed satisfied local
goals. The idea being to determine what local area needs attention now
to lead to desirable positions later (much later) in the game.   Once
the important areas are discovered standard tactics, patterns,
searching, etc. can be used to concentrate on playing well locally. 
Make any sense?

Matt