Sagarin ratings have been around for quite some time (I believe ~4 1/2 decades).
He's either a mathematician or statistician who applied a chess rankings algorithm (ELO) to sports. I believe he has his own algorithm and only uses ELO to compare outcomes but I could be wrong there.
I'd love to know how his stuff actually works because I've never figured it out and would like a method that I don't feel is just a slap together method.
As an active research scientist in this field i find it unlikely Sagarin has a method of analysis that is beyond me. Then its just a matter of what kind of groundings seemed "good enough" to him. Same with Massey.
BTW, Sagarin and massey have more than one measure with their rankings, some score based, some not. Now I don't know how they try to predict game spread or other things.
We more or less know that their methods are some form of binary regression with some fudging. How much of it is like Kalman filter where you advance previous inputs? How much is it like maximum-likelihood? Both are kind of "I don't know".
If this is above peoples heads... fine, it is what it is. I've wanted a method of my own since I entered grad school and i'm here 18 years later with nothing to show for it. I could create something that's kind of a hodge-podge but I wouldn't believe in it.
That being said, the rankings have the advantage in that they can compare all the results others aren't willing to compare. This has the biggest impact in college basketball when certain conferences have up years. We're also in early-season football. You technically can't compare teams that haven't played each other through a chain of games. They're just flat out unrelated. So you're then making statements based on past information or blatant fudging. I don't feel like rankings can even really make sense until 6 or 7 games in.
The ideal of rankings is it is able to process all events under a blatant supposition that "only wins matter" or "only score matters" and all other nuances are lost for good or for bad. They aren't perfect, they don't have secret knowledge. Ideally the ranking is just a balance of their competition and their win rate. You're below most of your losses and above most of your wins in the balancing act.
I'll shut up now. I know way too much about this stuff and enough to know they're all limited. We suck, we know we suck, is it 170 vs. 160? Well, that's why the play the games. Only way to find out.
