You can now find full, updated 2017 S&P+ rankings at Football Outsiders — overall, offense, defense.

What changes have I made?

Instead of filling space with the text here again, I’ll just direct you to the other posts in this series.

The short version:

1. I changed the garbage time definition.
2. I increased the weight and longevity of preseason projections
3. I introduced ways to make S&P+ more reactive to actual game results.
4. I incorporated conference-level adjustments to account for occasional connectivity issues.

For further details, click on any of these posts:

On with the 2017 rankings!

Return of the P5

In walking through these rankings, season to season, it’s been interesting seeing the relationship between power conferences and mid-majors expand and contract like a lung. The 2016 season saw a contraction, with three of five P5 conferences seeing their S&P+ averages shrink (the SEC by a lot) and four of five G5 conferences rising.

In 2017 came the opposite. Four of five P5s improved, and all five G5s regressed. And the SEC, which had reached a high point in 2014 before stumbling dramatically, began a rebound that would pick up a lot of steam in 2018.

Full-season average S&P+ ratings, 2017:

1. SEC (+12.8, up 2.6 adjusted points per game from 2016)
2. ACC (+11.3, up 1.0)
3. Big Ten (+10.0, up 3.1)
4. Big 12 (+8.7, up 0.4)
5. Pac-12 (+7.0, down 1.0)
6. AAC (-1.6, down 1.8)
7. Mountain West (-7.1, down 2.4)
8. MAC (-7.4, down 3.1)
9. Sun Belt (-10.3, down 1.4)
10. Conference USA (-10.4, down 2.1)

The ratings distribution overall was kind of interesting in this season: the top 20 teams all ended up with higher percentile ratings than average for their ranking. Then, starting with No. 21 Virginia Tech, the percentiles quickly became worse than average.

Let’s put that a slightly different way: In 2010, the No. 1 team in S&P+ (Alabama, of course) had a percentile rating of 96.9 percent. In 2017, the No. 5 team (Clemson) was 97.0 percent. Any of 2017’s Alabama, Ohio State, Wisconsin, Georgia, or Clemson would have been the top-ranked team in 2010.

But anyway, we’ll put this in the Things That Interest Only Me category and move on.

Or not:

Wisconsin’s presence ... is kind of infuriating, actually

Photo by Mike Ehrmann/Getty Images

I spent a lot of 2017 writing about how that year’s version of Wisconsin was different than others. Or, more specifically, Wisconsin’s quarterback was different than most Wisconsin quarterbacks.

The Badgers had an insanely good defense and their customary high-volume rusher (freshman Jonathan Taylor went for 1,977 yards), but they could also pass when they needed to. On third downs with between 4 and 9 yards to go, Alex Hornibrook boasted a 158.8 passer rating. They stayed on schedule, and when they didn’t, Hornibrook bailed them out far more than he got credit for.

Wisconsin went 13-1, losing only once — by six in the Big Ten Championship to the second-ranked team in S&P+, Ohio State — and beating four other top-40 teams by an average score of 32-18. The Badgers were legitimately awesome; they weren’t just a team racking up a gaudy record in a bad division, like Iowa had a couple of years earlier.

These stats provide some affirmation for what I found myself believing in 2017. They also make the Badgers’ — and Hornibrook’s, in particular — 2018 struggles even more frustrating. Hornibrook’s 2018 passer rating in that same third-and-4-to-9 range last fall was 122.0. He was hurt and missed a lot of time, sure, but aside from a brilliant performance against Iowa (17-for-22 for 205 yards and three scores), his pre-injury stats were already wavering.

Hornibrook missed most of the last half of the season, then transferred to Florida State. He, and Wisconsin, will always have 2017, at least.

Cruelty to UCF

Photo by Streeter Lecka/Getty Images

You know how I mentioned that the top 20 all graded out better than usual? Well, your only 2017 unbeaten came in ... at 22nd.

My new S&P+ formula more heavily weights recent history/priors by keeping preseason projections included in the projections for the full season. And the conference-level adjustments I added had the effect of slapping down the ratings of most teams from mid-major conferences while propping up a lot of P5s.

Once again, these make S&P+ more accurate and predictive as a whole. I keep wanting to remind people of that; this wasn’t some conscious decision to mean, and it wasn’t about aesthetics. But it did affix a lower ceiling onto the ratings of fast risers — teams that went from nothing to something almost overnight — and it lowered the ceiling of most mid-major programs as well.

In other words, it was almost like it was custom-built to stomp on UCF’s 2017 team. It wasn’t, I promise. But it sure looks that way.

UCF, of course, went 13-0 that season, laying waste to most of the AAC (and Maryland!) during the regular season, surviving a classic with USF (S&P+ No. 32) and Memphis (No. 42) late, then eking out a win over Auburn (No. 8!) despite getting out-gained in the Peach Bowl. I believe the original S&P+ formula had the Knights finishing at eighth or ninth, and on pure achievement that sounds right. But S&P+ is intended to be predictive, even at the end of the season, so these priors- and conference-adjusted pieces stay in the formula even at the end. Sorry, Knights.