Mark E. Glickman

And then, you know, fans are gonna keep voting every day throughout the entire year.

So how much weight should you give to a vote in week one versus a weight to a vote at the end of the year? Uh, and you know, you can think of like a naive way of doing that, of going, "Oh, one week of games, it's worth one, two weeks of games, we're two." But you know, I'm a statistician, so I, uh, I took some time and, uh, was reading, of course, a classic work by Mark Glickman and also Shane.

There's this cool paper of thinking about, you know, how much infor- They were looking, I think it was like for design of matchups and selecting, like how much information do we gain-

Pair comparison, yeah.

Yeah, in the pair comparison test, right? In pair comparison modeling.

And so what I, what I decided to do was, you know, we get, we-- In college football, there is very Limited, I would say, like public available college football stats to work with, right? I don't have statistics yet on every single position to say like, how much information have I learned about every player in college football? But what I do have is I have game outcomes.

And so I can model head-to-head outcomes of games, and this is, uh, going back to classic Mark Glickman, Howe Stern model for football outcomes, and I have strength parameters that I learn for each team, and I can allow that i- to evolve from week to week, as well as from season to season.

And what I, what I ended up doing was I was curious about like, okay, how much information am I gaining? You know, how much information am I observing from week to week? And so what I decided to set up is if I viewed the, the ratings I effectively learn from this state space model at the end of the year as sort of the anchor, I could see how far away are we from that end of the year at week 12, at week 11, at week 10, and so on, all the way back to week one.