Why is FiveThirtyEight’s college football playoff odds calculator so absurdly bad?

I realize this is a trivial subject in and of itself, but given that Nate Silver’s site has an up till now well-deserved reputation for predicting decidedly non-trivial political outcomes, the fact that its college football playoff calculator is laughably dysfunctional may be of interest beyond the world of sports fans.

The four teams that will be selected for the college football playoff will be chosen by a committee.  This means that any calculator is predicting the voting behavior of a group of human beings, not the outcome of some algorithmic computer power ranking or the like. And here are a few example of what the FiveThirtyEight calculator is predicting that group of human beings will do:  (BTW I’m not commenting on the substantive merits of the predicted decisions, just on the likelihood of what those decisions will be).

(1) If Notre Dame, Ohio State, and Oklahoma all win out, the calculator puts Ohio State’s chances of making the playoff higher than Notre Dame’s (75% to 68%), while Oklahoma’s chances are only slightly lower (66%).

This is utterly absurd. If you know anything at all about how college football works, you realize the chances of an undefeated Notre Dame being beat out for a playoff spot by a one-loss OSU team (whose loss was a rout by a mediocre Purdue team) are zero. Z-E-R-O.  That will not happen.  The Oklahoma over ND scenario is similarly bizarre.

(2) If Alabama, Clemson, and Michigan win out, the calculator puts Michigan’s chances of making the playoff as 78%.   If all that happens, Michigan’s chances are actually 100%.  The only conceivable scenario under which Michigan wins out and might not make it is if Alabama’s only loss is to a one-loss Georgia team in the SEC championship game.  Again, that’s just how this stuff works.  I can’t program a computer to spit out that result but I know it in the same way I know that Jill Stein isn’t going to be elected president in 2020.

(3) Along similar lines, the calculator gives Ohio State a far better chance (91%) of making it if the Bucks win out along with Alabama and Clemson than it gives Michigan of making it if Michigan wins out along with the two southern schools (78%).   This, too, is just completely ridiculous.  In these two respective scenarios, Ohio State’s loss would be far worse than Michigan’s (a one-TD loss on the road to undefeated Notre Dame.)

There are plenty of other major malfunctions along these lines. (If you’re a college football fan, you can play around with the model yourself at the link above).  I would be curious to know how Silver’s team came up with its model and maybe I’ll try to ask him.  Anyway it’s just a mess.