Louisville is the argument
Fifteen spots is the widest gap on the board. The voters have Louisville 24th. We have them 39th, which is not a top-25 team in any reading of our numbers.
The disagreement is about the schedule more than the roster. Our model reads a path that produces a lot of ways to finish 8-4, and it does not hand out credit for a close loss. The poll is doing what preseason polls do, which is start a good program near where it finished.
If you follow Louisville you already know which way you lean on that. One of us is wrong and the season will say which, which is the whole reason this number goes up in August instead of November.
Where we are higher than the room
Michigan is our biggest bet in the other direction. We have them 7th and the voters have them 16th, a nine-spot gap that comes almost entirely from returning production and a schedule our model likes more than the room does.
USC lands 8th for us against 14th in the poll. Houston is 18th for us and 23rd for them. Notre Dame is the loudest one, because we have them first overall and the voters have them fourth.
Three teams the poll left out sit inside our top 25 anyway. Florida 20th, Auburn 24th, Arizona 25th. If you want somewhere to point and laugh in October, start with those three, because they are the calls most likely to age badly and they are on the record now.
Where the room is higher than us
Georgia at 3rd and Ole Miss at 9th are both six spots above where we have them. Utah is 21st for the voters and 27th for us.
You can probably see the pattern without being told. The teams the poll likes better tend to be programs with a decade of brand behind them. Nobody is accusing the voters of doing that on purpose. The point is that our model has no idea who any of these teams are, and the gaps land in the same places every year.
The number we used, and the one we did not
This comparison uses our model-only ranking. Power, schedule path and expected wins, with no poll input of any kind.
That matters this week because we just started feeding the real AP and Coaches polls into another part of the system, the layer that models what the selection committee is likely to think. Our main board deliberately blends our model with public perception, so the board now contains a little of the AP inside it.
Ranking this page that way would have compared the poll against a number partly made of the poll. It shrank Michigan's gap from nine spots to six before we caught it. Every gap above is measured against the version of our rating that has never seen a ballot.
December settles it
None of this is worth anything as an opinion. You can get an opinion anywhere. What you cannot get anywhere is a number stamped in August, in public, graded in the same place every week whether it looks good or not.
If Louisville makes the playoff, that 39 goes on the Ledger next to the result and stays there. If Michigan wins the Big Ten, you will hear about this post. Either way you get to check the work, which is the only thing here worth your time.