Last year I thought the ESPN FPI underestimated Sun Belt team's chances against non-conference teams so I decided to come up with my own spreads to see how they compare to the FPI.
What I did is take the
preseason S&P rankings from FBS and use each team's performance for 2018 and 2019 to extrapolate how they should do in 2020.
One thing I'm accounting for that I doubt most computer predictions don't is consistency. Teams that were more inconsistent in 2018 and 2019 will have a better chance of losing to a team they're favored over than more consistent teams will with the same spread and a better chance of an upset as an underdog than a more consistent team with the same spread.
Think of how ULM played Georgia State and Florida State last year versus how they played App State and Iowa State. Also compare how GS played Minnesota and App State versus how they played UL and Liberty. That's what I'm trying to account for, here. Of course the better teams from 2018 and 2019 will still have more favorable spreads against any one team.
GS, Troy, and ULM were the 3 most inconsistent teams and Texas State, App State, and Louisiana were the most consistent teams.
Here's the spread for FPI and what I came up with:
Notice how App favored over Wake and Coastal favored over Kansas in FPI. I can't remember ever seeing a SBC team favored over a P5. The FPI likes Coastal more than my model didn't and doesn't like ULM nearly as much. Other than that, as much difference as I thought.
BTW, I would expect another year with several P5 wins. The FPI is expecting 3, my model is expecting 4. I could see any of those games other than Alabama, Georgia, Michigan, and Florida being wins.