I don't actually know how much weight it's giving to the prior season at this point in the season. I do know that FPI becomes less accurate the further into the season we go, but I think that's more because later season games are harder to predict than early season games. For example, it's harder to predict how BYU will do against UCF than SIU.

But if we assume it is giving weight to the prior season, this is because all of the work that went in to build the model showed that this was predictive. And now we can take the model results and see that they are doing quite well overall. It missed pretty big on BYU this year. Or at least it seems to have. It's really hard to tell with such a small sample.

I actually went and did the work to look at this objectively. What part of my analysis I posted above do you disagree with?