On Tue, 19 Aug 2003 14:08:33 +0100, "Sid" <[EMAIL PROTECTED]> wrote:
> > I'm sorry if I'm sounding like I'm discouraging you, Sid, but I am
> > trying to discourage you. Your task is extremely complex, if you're
> > talking about sports I am familiar with; you yourself will have to find
> > the right weighting scheme out of a zillion that exist, we cannot do
> > that for you; the goals you have set out (detect a 10% difference
> > between methods) are extremely tough to achieve; getting accurate
> > predictions is also extremely difficult to achieve. Your goal is
> > admirable, your path is difficult.
>
>
> Fair enough. But I still think it is a reasonable goal to be able to "rank"
> teams or individuals.
> I was not asking for the "right" weighting scheme. I was just asking for a
> few pointers.
Now you are making a 90-degree turn in what you
"are just asking."
Rating "performance for the season" is easier, since
it makes little or no use of which events came first.
You started out trying to compute a running
average with exponential-weighting for "decay";
the complication being, you had two variables for
the passage of time ("days" or "games").
Do you want adaptive-rating systems, ones that
make best use of competitive results? One of the best
seems to be the system used in international chess.
I think that one is documented for the public, too, so it
is probably available online.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
"Taxes are the price we pay for civilization." Justice Holmes.
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
