Zeke5123 wrote:
FromPittWithLove wrote:
Basic forecasting methods? You are trying to reduce every multivariable football decisions into Bayesian statistics? I'm assuming you've looked at the structure of Bayes theorem, please explain how you would reduce this particular situation to P(A|B) = P(B|A)*P(A)/P(B). I'm really interested in your definition of P(A) and P(B) since priors are generally really difficult to determine without thousands of samples (usually why most empirical studies simulate these parameters to determine the best fit of their data). .
Football is not economics (I'm guessing that's what you work in - most scientists don't use the word forecast but I apologize if I'm off) because the variables can't be reduced the way you are describing. Maybe in terms of FG decision making but even that's not going to be a very flexible model.
The same argument above holds true for just about every modelling avenue I'm aware of - so even if this isn't about Bayes Theorem I struggle to see how we could possibly measure MT with the metric you propose.
You are taking it to literal. I believe I said Bayesian reasoning. The basic idea, which you are obviously familiar with, is that there are two things that can be true -- in the case, the possibilities are (1) James Harrison is currently better than Moats or (2) Moats is better than Harrison.
Our prior is that players that are border line HOFers are better than guys who have always been scrubs. Thus, probabilistically we can say state 1 is more likely than state 2.
The next issue is there any recent evidence that suggests the prior is wrong in this case. The relevant data is that very recently Harrison performed at a starter level, and has seemingly been better than Moats in limited snaps this season (both guys have only played somewhat limited snaps, but Harrison is the only one to make a splash play). The final piece of evidence is the decision by the coaches. I just don't weigh this evidence very highly. Coaches may believe the difference between Harrison and Moats is rather marginal, and the coaches want to develop Moats for next year. Or, maybe the coaches believe Moats is the better player. Or maybe politics played into it -- wouldn't be the first time. Or, maybe the coaches got it wrong. Regardless, don't think that decision will make me change my belief that state 1 is more likely than state 2, though it does decrease the gap.
In the same way, the fact that BB is interested in Harrison also influences the probability. Now, BB didn't have the choice of Moats or Harrison; thus, the inference from this is limited. But count this as some evidence towards state 1.
Thus, applied basic Bayesian reasoning (not the Bayes theorem) to football decision. Not that hard.
And you got me -- economics undergrad. Went to law school though and became a tax lawyer. New tax law is very interesting.
But Zeke, you're leaving out a very key point
Moats played 96 FUCKING SNAPS on special teams this year!!
How the fuck could we EVER replace that type of unbelievable football production??
Harrison might be a HOF outside LBer, but WHAT ABOUT SPECIAL TEAMS!!!
Without Moats, we'd probably give up a TD on every kickoff, because then we'd be FORCED TO PLAY WITH ONLY 10 GUYS -- because surely, there is NO ONE ELSE on the roster who can take over Moats' fucking role on special teams!
Without Moats, our season WOULD BE FUCKING DOOMED!!!