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

Re: computer-go: extracting date to beat 9d from chess and draughtsscene

At 11:30 AM 11/14/99 +0100, you wrote:

At 12:32 AM 11/14/99 -0800, you wrote:
>At 12:20 AM 11/14/99 +0100, you wrote:
...
>>Now as we all know GO because of branching factor
>>is a bit harder than this.
>
>its a *lot*harder. in the late fuskei and early midgame. locally you can
>easily have 20-30 open spaces, so you are looking at say
>25! ...

For the 20-30 open spaces, the formula is not 25!

With alfabeta pruning that's already reduced to around squareroot( 25! ),
though that's still a huge number...

not that familiar with ab, but i'll take your word for it.


Then applying nullmove ...

don't know what nullmove is


...
>>Apart from that we also need a couple of tens of years before
>>Go programs get at tactical decent levels.
>
>maybe much sooner, tactical stuff can probably done fairly well...
the first problem will be to choose the right tactics. ...

That is shifting the problem from search to intelligent selecting moves,
which is dead wrong, and can be proven wrong directly; if you already know
what move to select say at 9d level, then a say 15 years old would already
start at 9dan level after taking a good look at how your source code selects
moves!
we don't know the move, but we could know the goal (i.e. connect my stones,
make a live shape).

>>Most important thing now might be to estimate how deep ...
>i am an amatuer 1-dan. i can look ...
>depending how much of a one-way street it is...

Right, let's skip one way streets, as those are very simple to travel
for both mankind and computer.
no, a sort of worst case example is the 17 stone sacrifice. this is a
one-way street, but only to a human that can see it. the machine will have
to search, and since its a long sequence, will probably not find it, unless
it maybe had some goal directed smarts helping it out.

The difference between pro's and amateurs is not the number of moves
they look ahead. De Groot already proved that the basic difference is
the quality of what they see...
don't know him, but i doubt it's a proof in the mathematical sense. could
you point his stuff?

...
>you won't get that far in an open area of the board (say 30 empty spaces).
>you can do real well in the endgame. and nothing at all in the fuseki.

Are we talking here about local tactical search?
not sure what you mean here. i was thinking of 15-40 stones on the board
(late fuseki or early midgame).

let me try to clarify my thinking. even assuming moor's law continues and
we have machines that are say 10^6 times as fast as we have today in say 20
years. you still have no chance in the late fuseki and early endgame. if
you look at a completed game that was fairly peaceful, the winner is the
one whose stones are the most efficient in surrounding territory. the pro
will simply find a way to make his stones more efficient than the
computer's. this is fairly clear to me, but i don't know if i have
explained it very well.

thanks


Ray (will hack java for food) http://home.pacbell.net/rtayek/
hate spam? http://www.blighty.com/products/spade/