Re: new ranking calculation

On Thu, Oct 09, 2003 at 08:05:42PM +0200, Horn G=E1bor wrote:
> Hi!
>=20
> A quote from the original gamasutra article:
>=20
> It is usually not a good idea to combine ratings from different game
> modes. Teamplay has different objectives, and mixing scores can only
> lead to confusion. Performance statistics based on simple criteria
> (hits/shots, damage done vs. damage received, average lifetime) work
> best, and largely avoid credit assignment issues. Rating groups of
> players as a team is closer to the information needed to select player
> team pairings.
>=20
> (well it was for quake)
>=20
> Please forget the first simply summarizing method, that's obviously not
> good, was more than a probe for me (i got the doranking script yesterda=
y
> and it's the first time i try perl) than a solution.
>=20
> I tried to shape the new formula due to mathematical base. Please let's
> concentrate on extendending/fixing that.
>=20
> U mentioned ELO can be extended for multi-player. I didn't say it
> cannot, i said i didn't find a way to extend it to be suitable in MIXED
> environment! Gamas articel also said that.

Well, not to get too far out there, but does it have to be mixed?

I don't have a problem if players have more than one score.  (Battle.net
does this for different type games, FFA, team, 1v1).

> So it's a good basis, but we need a special solution here (but of cours=
e
> if u can point to any existing formula, that would be very good).
>=20
> bye, hirisov
>=20
> > Hmm.  The modifications looked significant and non-mathematically-sou=
nd,=20
> > although I haven't studied closely.  I know that ELO can be extended =
to=20
> > multi-player in a mathematically sound way (I have done this before;=20
> > http://ggz.sf.net/), but it requires going back to the original=20
> > principles and working things out from scratch.
> >=20
>=20

--=20
Paul Zastoupil