Re: [computer-go] Pattern matching - example play

["Frank de Groot" <frank@xxxxxxxxxxxxx>]

From: "Persson, Magnus" <persson@xxxxxxxxxxxxxxxxx>
Subject: RE: [computer-go] Pattern matching - example play



>I reached dan-level mainly by studying books and pro games, and played a
>tournament about every 3 or 4 months.

You are in fact corroborating my postulation that a program can learn from
observed games..
Books contain positions + "good moves" and observing pro games, well, that's
what I am advocating a learning program to do..


>Trying to understand moves that you do not understand at all is the key
>to becoming a better player.

A learning computer program does this too, but differently from a human
being.
It discovers relationships and values.


>place when you try use new concepts or try to mimic the shapes played by
>better players and see the consequences. No go program I know of is
>truly capable of doing this.

Well I did not advocate building AI, merely learning good shape, eye shapes,
connectivity etc.
My claim was that a learning system does that better than a Go playing
programmer.



>Perfect shape predictions is also fine but will not solve the problem of
>evaluating the whole board.


*my* shape system always evaluates the entire board.
All my shape values are based on a board with infinite size and all patterns
are on the board simultaneously.

But of course a pure "shape" system doesn't see ladders (reliably) etc.
And it doesn't see tactical stuff deeper than 1, 3 or perhaps max. 5 ply.

So the issue is not that learnig systems suck or pattern systems suck, the
issue is that we need to "learn" more aspects of the game and hybridize it
with search and applying the rules of Go in that search.

_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/