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

Re: computer-go: perfect players


Note that the handicap between the current and past Gnugo, is probably not the same
as the rating change from one to the other. Since computer go style is different from
human style, differences in strength between programs is often not reflected in
differences in games against people.

I think that improvement among the top programs is less than 1 stone a year. Gnugo
is improving faster than that right now, but is still catching up to the best programs.

At 09:36 PM 5/4/2001 +0200, you wrote:
On Thu, May 03, 2001 at 07:07:53PM +0100, Jon Diamond wrote:
> What I wonder is how quickly are the top programs improving?
> Is there anyone who can suggest a graph over time?

I have recently measured gnugo 2.6 (feb-2000) agaisnt the latest 2.7.233
(Apr 29, 2001), and found the difference to be between 4 and 5 stones (4
stones: 2.6 won 55% of the games; 5 stones: it won 45% - averages of 100
games series).


David Fotland