Originally Posted by mdgator05
Unless I saw some actual stats disproving it, I would be very surprised if there wasn't state dependence in this data.
Think about it like a random walk. So maybe every team is subject to the same normal distribution of playing better or worse on a given day, with a mean at where they played the last game and some standard error set evenly for all teams.
Even in this year, I would be shocked if there wasn't state dependence in this data. It is a pretty crucial component. Basically, if you took your predicted model and found significant heteroskedasticity, in which, to use Florida as an example, they consistently outperformed the projection early in the year versus underperforming the projection later in the year, you would have a pretty good argument that model accuracy could be improved by the inclusion of some measure of state dependence.
Not meaning this as an attack as I appreciate the use of stats to analyze teams above fairly biased observations as given by the talking heads. Just thinking that might help model accuracy if it isn't there. And if it is, that is a pretty surprising result.
I appreciate the logical response, and I understand your point a little better now, but two things:
1: I think Florida has been consistent this season. Early on when we played bad opponents (relative to location) we destroyed (easy teams, good teams (Marquette, Wisconsin at home)), and we struggled against good teams (relative to location again) (Kansas St. on road Arizona on road). Same could be said for late season favorable scenarios we handled easily, unfavorable ones we struggled. Heck we seem the MOSt consistent if you think about it (maybe not the best team however,
2: Even if that analysis were incorporated into the calculation I just have the opinion that it would not affect the ratings too much.
The point of this thread guys was just a general outlook to keep the entire season in perspective. I do not believe we are top 5, just that we have a GREAT shot at doing very well.