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

Re: [computer-go] Search = Bad!



John Tromp wrote:
I have computed the game theoretical results of all positions in the game
of connect-4 as played on a 7x6 board, with up to 4 stones. The way I map
this depth-4 game tree into 2 dimensions is to start with a big circle for
the root, and then draw the 7 1-ply positions as subcircles of that one,
and the 7x7=49 2-ply positions as subsubcircles etc. Each circle is colored
in one of 3 shades, represting a loss, draw, or win (but then partly obscured by the smaller circles drawn on top of it).
Ok, exactly what I thought. Just depicting the game tree in 2 dimensions. Its a very nice picture though ;)

Apparently this is not what Frank is trying to describe however, because when I said as much, this is what Frank replied:

Frank de Groot wrote:
To me it seems like what your describing would simply be the game tree represented in a 2 dimensional form.
Have you read all my postings?
To me it seems you have not read anything I wrote.

Listen, I am logging off until 2006.
Nice. So he either means the statement was true, but so obvious its not worth restating, or its false and his postings indicate something else which I/we am/are not able to discern.

John Tromp wrote:
I was simply reminded of my cover when Frank mentioned fractals, because
that's how always talked about it to people: "look at this graphical
representation of early connect-4 positions; isn't it fractal like?"
I understand.
But I'm afraid the shortest formula to describe this fractal would be
the brute force game tree search:(
Precisely.

For go, a shortcut formula is even less likely, since its PSPACE completeness implies that any fast (subexponential) algorithm for go would translate to
a fast algorithm for *any* problem whatsoever solvable in polynomial space.
Agreed.

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