Jeff Sagarin - USA Today

[Peregrine Stoop]

use sagarin
adjust for turnovers
adjust for special teams
adjust for 4th down and 3rd down luck

profit

[aggieshawn]

do i sleep with an old chick tonight? hurrry

[dodger33]

do i sleep with an old chick tonight? hurrry

Is she fat?

[brettd]

The second continuum is based mostly on the (non-data-mined) sample size behind it. This is something that you should basically be thinking of in terms of Kelly Fraction; that is, as your confidence in your model increases, so should your Kelly Fraction. You might want to write an equation translating z-scores into Kelly fractions (search this sub-forum for more info on z-scores or Kelly).

This is something I'm wrestling with at the moment. How do you quantify 'confidence' quantitatively? I'm sure each person has their own way, but I just want to know how other people come up with their 'confidence continuum'.

[evo34]

use sagarin adjust for turnovers adjust for special teams adjust for 4th down and 3rd down luck profit

What do you mean by, "adjust for special teams?" Obviously you need to neutralize anything you think is more luck than skill and has influenced past scoring (turnovers and 3rd down rates, like you say). But I do not think special teams (other than excessive returns for TD) is mostly luck, and thus should not be adjusted for directly when tweaking a score-based power rating system. Do you disagree? If so, what kind of adjustments do you make?