ESPN's FPI is a pretty slick rating system with sound mechanics behind it. Too bad their own writers don't quite know how to interpret its results.
According to the Allstate Playoff Predictor, Ohio State still has the best chance to win the Big Ten (37%), followed by Iowa (31%), Michigan (17%) and Penn State (8%). The Buckeyes are also still favored to win the Big Ten East (51%), followed by Michigan (26%). There are plenty of opportunities for Ohio State to compensate for its loss to Oregon, and right now, ESPN's Football Power Index projects the Buckeyes to win all of them. -- Heather Dinich
There was more, from other writers saying that Ohio State has been playing the best football in the Big Ten. but I can't find it right now, Doesn't matter. The upshot is that you can't say a team is "coming on strong" if they're ranked high in the FPI if that team was ranked high in the FPI pre-season.
The FPI is still laden with pre-season "priors" to make it more accurate at this early stage. So part of Ohio State's high rank is, quite simply, bias from last year.
Above: ESPN FPI top three - games thru October 2, 2021
With Ohio State at #3, it makes sense that they'd be favored to win the Big Ten over the likes of Iowa, Penn State, Michigan, et al.
But if you look at the "Efficiencies" tab of the FPI, you get the real story—how teams are objectively doing in THIS YEAR'S GAMES only. No watering down with pre-season projections.
Above: ESPN FPI Efficiencies - game thru October 2, 2021
WHOA. Suddenly Michigan is #3, well ahead of #11 Ohio State. And #3 Iowa and #9 Michigan State also have played better this season so far. #12 Penn State is just one spot away, too. That's a lot of teams for Ohio State to surpass on their way to a title.
Which is more accurate? Well, early on there's no question that ratings with pre-season bias are better—there's only a game or two to go by, and no worthwhile conclusions can be drawn. But the farther you go, the more you have to start to consider the "pure" ratings until, at some point, you ween the ratings of any pre-season bias.
That will happen in a couple of weeks. Right now, there is still value in using last year's info in the form of pre-season ratings. But the "pure" ratings reflected in the Efficiencies tab is starting to become pretty solid, and in many cases might reflect reality better than the biased FPI.
Our own ratings have this same phenomenon. Our "Bayesian" ratings, which we post from time to time, are modified pre-season ratings. The Strength power rating on the other hand has no bias. A few weeks ago the Strength power rating looked pretty silly in a lot of ways, but after 5 weeks it's coming together nicely. Not coincidentally, our top teams in the Bayesian rating this week are:
- Georgia
- Alabama
- Iowa
- Ohio State
Very similar to the FPI. And in Strength? Our current Strength power rating looks like this:
# Team rec Rating Offense[rnk] Defense[rnk] FPI Efficiency: 1. Georgia 5-0 89.93 73.35 [ 2] -16.57 [ 3] 1 2. Michigan 5-0 82.86 71.28 [ 4] -11.58 [ 6] 3 3. Alabama 5-0 79.86 71.85 [ 3] -8.00 [ 17] 2 4. Pittsburgh 4-1 79.58 78.17 [ 1] -1.41 [ 56] 6 5. Iowa 5-0 79.34 62.35 [ 17] -16.98 [ 2] 4 6. Cincinnati 4-0 73.61 64.97 [ 9] -8.64 [ 14] 5 7. Iowa St 3-2 72.62 63.45 [ 13] -9.17 [ 11] 26 8. Florida 3-2 72.13 59.82 [ 26] -12.31 [ 5] 8 9. Ohio State 4-1 70.44 69.11 [ 5] -1.33 [ 57] 11 10. Mississippi 3-1 69.72 64.02 [ 11] -5.69 [ 31] 13
That's pretty close. Only Iowa State is off-script. Michigan and Iowa still outrank Ohio State so far this year, in an unbiased sense.
I'm pretty sure the mid-point of the season is when they ween the FPI from it's pre-season rankings, so In a few weeks the FPI and Efficiencies will reflect the same ratings. With two more games per team, roughly, we'll have enough information to make judgements without the help of the bias that keeps the early ratings looking good. And then we can say whether Ohio State is really the favorite to win the Big Ten. Because right now, using no bias, they're not.
Comments