ESPN released its preseason FPI rankings and the system is not high on Florida for 2021. After the Gators finished 11th in FPI last year, they fall to 14th in the preseason edition.
FPI projects an 8-4 record — 8.2 wins and 3.9 losses to be precise, which tells me there’s something funky going on with the rounding. It gives UF a 97.3% chance going to a bowl, a 17.2% chance to win the East division, a 3.9% chance of winning the SEC, and 2.7% chance of making the Playoff.
While surveying a few different preseason computer projections at an Oklahoma blog, I noticed that they’re all fairly down on Florida. The team falls from 6th in SP+ last year to 12th in the preseason, and in Sharp College Football it falls from 11th to 17th. The fourth model listed is not one I’m familiar with and is paywalled, so I don’t know where UF ended last year. It has UF at 16th.
The writer of the blog post averaged out the rankings, and Florida actually did better. The Gators ended up tied with Wisconsin at 10th, even though their average rank was 14.75. What that should tell you is there is a lot of disagreement among the systems.
FPI is a marginal candidate for a serious listing like this. If it wasn’t backed by ESPN, it would be laughed at for its, let’s say eccentricity. It has Mississippi State at 8th, for instance, and Oklahoma State at 9th. The absurdity of the former is self-evident; the latter team appears only in one other system’s top 25, at 20th.
Beyond that, Miami ranges from No. 8 (SP+) to unranked (Sharp). Washington ranges from No. 11 (SP+) to unranked (FPI). Texas is in two top tens but is also No. 20 elsewhere. Notre Dame is in the top 11 twice and in the 20s twice.
I wrote before the 2018 and 2019 seasons why these kinds of systems might undersell Florida. I think there’s a chance of it happening again, and here’s why.
Returning production
Bill Connelly, who runs the SP+ system, has done a couple of podcast interviews recently. In one of them, he mentioned that a lot of teams have an unusually high amount of returning production. His SP+ rankings in the post listed above were initial and come from two months ago. A lot of teams, especially in the G5 leagues, didn’t have completely solid information about who is coming back because of the uncertainty with the COVID eligibility mulligan.
He said that information is much easier to come by now, and he’ll put out an update to his preseason SP+ later to reflect it. I wouldn’t be surprised if the Gators fell after the update. Connelly said that the average returning production this year is about 70%, which normally would get a team into the top 30. A ton of teams have returning quarterbacks, specifically.
Florida lost Kyle Trask to the NFL, so that was always going to be a big hit. But Kyle Pitts also went pro, as did Kadarius Toney, Trevon Grimes, and a lot of the Gators’ secondary. In Connelly’s initial rankings from February, UF was 108th in returning production. Sharp has the Gators at 111th.
The good systems like SP+ do account for transfers in returning production, but UF doesn’t get a lot of help there. Demarkcus Bowman did little for Clemson a year ago, and defensive tackles’ value isn’t well captured by numbers.
UF didn’t have a mass exodus like after the 2006 or 2009 seasons. It was a normal amount of players leaving with few super seniors taking that eligibility do-over. It’s just that in 2021, having a normal amount of departures means you have a relatively small amount of production coming back.
Recruiting rankings
Because Dan Mullen has leaned on the transfer portal, the Gators do not have as high of recruiting rankings as they otherwise might. I believe pretty much all projection systems use four-year recruiting class rankings with some kinds of weights, and I’m not sure if any use a total team talent rating like what 247 has. As a result, Florida will show up a little lower than it should in the pure talent part of these ratings.
Now, it’s hard to get these things right anyway. What should Lorenzo Lingard go down as? He was a 5-star recruit, but he went through significant injuries at Miami and was still recovering to a degree last year in Gainesville. He probably shouldn’t be a 5-star anymore because of those issues, but the spring practice reports for him were glowing. If he can get back to his old self, maybe he should be a 5-star.
I don’t think Justin Shorter will ever live up to his 5-star billing, though I think he can do more than he did last year with more time in the offense and without guys like Pitts, Toney and Grimes ahead of him. Antonio Shelton has already outperformed his modest 3-star high school rating.
UF still has been in or just outside the top ten of the recruiting rankings, so it’s not like they’re being majorly downgraded here. But when margins can be slim, this could play a factor.
The defense
Florida’s defense was bad by its standards last year. There was bad strategy going on, something that was fixed by bringing in a couple of new coaches and changing up coverage styles in the spring. Thinness at defensive tackle hurt the back seven, but stats don’t have a way to model that for the same reason they don’t have a great way to model Shelton and Daquan Newkirk solving a lot of that specific problem.
The Gators should have a much improved defense so long as they don’t lose one of the top three corners to injury. Yet, these stat systems see a fall in defensive performance last year and a low-ish returning production number for the reasons discussed above. That’s a formula for underrating a team some.
The bowl game
Florida basically took the Cotton Bowl as an early spring game. Oklahoma did not, and UF got smoked. The numerical systems have no way to account for that fact, so 8.3% of Florida’s 2020 season is made of a really, really bad number from Arlington that is in no way representative of anything. Past performance is a component of these systems, with the most recent year weighted the most heavily.
The 2020 mess
I once wondered what it was like back in 1918 or the first half of the 1940s when teams played different numbers of games due to extenuating circumstances. I wonder no more because we just lived through it.
It is very, very hard for these stats systems to adjust for the varying sample sizes. Uniformity is a big help to them. So is connectivity between leagues, but there was very little of that with few non-conference games occurring.
The makers of these systems are doing the best they can, but they’ve never had a harder challenge. Because of small sample sizes most everywhere, anomalies can have outsized impacts.
The projections aren’t straight garbage now or anything, but they might be less useful than normal because of how messy the 2020 season was.