[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: computer-go: A little Arithmetic
In your estimation, how many games human has played, good plus not good?
If board configurations can be generalized, there should exit a learning
system that can generalize these configurations. The real odd balls can be
dealt by 9K programs. If board configuration can not be generalized, no
learning scheme will work.
I do have trouble to think which move is good or not good. A very primitive
thinking is to depend on statistics or information entropy. But that does
not sound right. That's the reason I was asking if some one here already
tried, and what the results are.
Weimin
P.S. I guess Mr. Cho will not like to play a 9K, or sub-standard move. A bad
move is a bad move.
----- Original Message -----
From: "Andre Engels" <engels@xxxxxxxxxxxxxxxxx>
To: <computer-go@xxxxxxxxxxxxxxxxx>
Sent: Wednesday, November 15, 2000 11:16 AM
Subject: Re: computer-go: A little Arithmetic
...
1,000,000 good games is a ridiculously low estimate. If we estimate that
there are 2 good moves in each position, with 250 moves we already have
10^25 possible games. 1,000,000 looks more correct as an estimate of the
number of ways the first 10 moves can be played good (according to the
standards of current professionals). On the other hand, of course several
of these games start the same way, so you have only 2.10^25 positions rather
than 2.10^27. On the other hand, in reality the number of good moves per
position will probably be greater than two on average. Especially since you
regard every move that would be played by someone stronger than 7 kyu as
'good'.
Some other objections to your proposals:
- How are you going to know which are the 'good games'?
- As soon as you play one 'not good' move, your system does not work
any more. If Cho Chikun plays your proposal and (either or not deliberate)
chooses to play slightly sub-standard on move 10, your computer #362 is
on its own.
--
Andre Engels, engels@xxxxxxxxxxxxxxxxx
telephone: +31-40-2474628 (work) +31-6-17774490 (mobile)
http://www.win.tue.nl/~engels/index_en.html