Hard question: How good statistically is the FPI model at accurately predicting matchups?

Easy question: How low is BYU ranked relative to where I, a blue-googled BYU fan, would like them to be ranked?

I think people are substituting the easy question for the hard one when then evaluate FPI. Yes, this year, the FPI model seemed to really miss on BYU. Based on the pre-season FPI numbers we had roughly a 1 in 1000 chance of going 10-2 in the regular season. So clearly the model underestimated us (although with 134 teams, a 1-in-1000 season will happen more often than you might expect). But the goal of FPI is to make predictions for 134 teams, not for BYU alone.

There is plenty of data out there to do your own analysis of FPI accuracy. Or you can rely on one of the sites that ranks CFB models. Or at least be consistent and complain about FPI equally loudly in the years it overrates BYU.

I do very much agree, to the degree that the playoff committee is relying on something like FPI alone for strength-of-schedule, they should not do that. Every model is going to have biases (in the statistical sense). They should be looking across several models to get a more wholistic view. But you could do far worse than relying on FPI alone based on how it has performed historically (including this season). In general a more model driven approach to rankings and CFP decisions would probably be an improvement. It's a lot easier to fix a biased model than a biased human (or committee of humans).