When I was organizing data for my look at the Pareto principle and recruiting rankings — the tl;dr version of that post: looking at the top talent on each team ends up being more predictive than any sort of multi-year recruiting averages — earlier this week, I went ahead and also pulled in size and experience data from my 2018 preview files. Stars matter, of course, but so do size/development and experience, right?

The former does, at least. The latter’s harder to figure out. We’ll save that for another time.

A quick summary of the process and data for this post:

1. For each year’s offseason preview series, I compile roster data, the previous year’s stats, recruiting rankings, etc., and maintain them in a public Google doc.
2. The roster data there is whatever the most updated info available is at the time I’m writing the preview. So, for the early-offseason previews for the Sun Belt, C-USA, etc., it’s probably just the previous year’s data (and the rosters probably don’t reflect every transfer and whatnot). For the final previews (the SEC and whichever of the other power conferences was second-best), the rosters have probably been more fully updated to account for both departures and updated heights/weights.
3. In these files, I basically list whoever I think it’s important to list. If you had a spot on the previous year’s two-deep or generated stats of some kind, you’re probably on the list. If you are an incoming freshman or a previous year’s redshirt that meets a certain threshold in terms of recruiting ranking (for Alabama, that cut-off might be that you’re a four-star recruit, for a Sun Belt team it might be more that you’re a high-two), you’re probably on the list. It’s a massively subjective process, but I usually end up listing the folks who will be taking the field in the fall. (I end up usually listing most of the players on the roster, for that matter.)
4. Because the thought is to potentially use this data in my preseason projections, I wanted to see what I could glean from these files, not from rosters updated in-season. So I pulled the roster data straight from the 2018 preview files. The correlations below would have likely been slightly stronger had I pulled in-season rosters, but that would have ruined the point of the exercise.

Using this extremely subjective and potentially flawed data, I still ended up finding some pretty impressive correlations between size and performance.

Here are the correlations between a given unit’s average size and a team’s S&P+ performance for 2018:

(I played around with height data and a BMI-style height-weight combination, too. Straight weight ended up with easily the strongest correlations.)

Backfield size doesn’t matter all that much, but offensive line size certainly seems to. And while size makes a bit of a difference on offense, it makes a much larger impact on defense.

Tactically, you can scheme your way to a strong offense. But with defense, size and recruiting rankings end up telling us quite a bit about what we need to know.

Now, one potentially obvious caveat: size and recruiting rankings are not isolated factors. We know a more fully developed offensive lineman (6’6, 305 pounds) is more likely to receive a four-star ranking than a 6’7, 245-pounder with a great frame and solid athleticism. We know that a 205-pound, carved-out-of-granite safety is going to draw the eyes of Rivals’ evaluators more than a 175-pounder with potential. So when we look at the data below, realize that I realize that.

Still, a lot of the ‘biggest’ teams in the country are teams that signed pretty big guys and made them really big. Let’s go league by league. Where did beef matter the most in 2018?

(Click on the images to enlarge them. There are a lot of numbers here, and there was no way to make this cleaner.)

SEC

League correlation between size and S&P+: 0.674.

To no one’s surprise, the SEC is the biggest conference in FBS — its 14 teams averaged 241 pounds per player, while the average for the rest of the P5 was about 236. That’s five extra pounds per player, or about 55 pounds per 11-player unit, for the conference that currently boasts eight of the top 19 teams in the country per S&P+. That’s obviously going to be good for the strength of the overall correlations.

One of the more interesting things to note here: two of the three biggest teams in the league were Mississippi State and Kentucky. Tennessee, meanwhile, was the third-smallest. While there are obvious ties between size and recruiting rankings, one list isn’t the other list. MSU and UK appear to recruit big guys pretty well and slap weight on others. You could throw Missouri in there, too — the Tigers were seventh in the league in S&P+ and eighth in size despite recruiting rankings that tend to fall around 12th or 13th in the SEC.

Big Ten

League correlation between size and S&P+: 0.365.

The four biggest teams in the Big Ten, as measured in the preseason: Penn State, Indiana, Minnesota, Wisconsin. Excellent, decent, decent, good. Northwestern was the smallest team in the league and won the West division (granted, while not necessarily being the best team in the West), and Michigan was the fourth-smallest and graded out as the best overall team per S&P+. So yeah, a blurrier picture here.

Northwestern’s really small, by the way, especially in the trenches. Among P5 teams, only Wake Forest had a smaller offensive line, and the defensive line was in the bottom 10 as well. So I guess size doesn’t always matter.

Big 12

League correlation between size and S&P+: 0.787.

In the Big 12, the five biggest teams were the five best. The six biggest offenses had an average Off. S&P+ of 38.1, and the four smallest were at 25.9. It was a different story on defense, where the three best units were third (Texas), fifth (TCU), and eighth (ISU) in size (and where Texas Tech had the biggest, and very much not the best, defense), but the overall correlations here were impressively strong.

ACC

League correlation between size and S&P+: 0.337.

For similarity’s sake, I included Notre Dame among ACC teams here. It made more sense than lumping them together with independents like NMSU and Liberty. Besides, noted ACC supporter Danny Kanell includes the Irish as an ACC team when it’s convenient, so why can’t I?

Broadly speaking, size mattered here. The eight biggest teams (including ND) boasted an average S&P+ rating of plus-9.7, while the seven smallest were at minus-1.2. FSU disappointed as a big team, and UVA, the second-smallest, still went 7-5. Still, you can make some broad generalizations here despite iffy overall correlations.

I do find it interesting that in a conference with three to four recruiting heavyweights (Clemson, ND, FSU, Miami), the biggest teams were NC State and Boston College. That very much suggests a concerted effort from Dave Doeren and Steve Addazio to develop size over speed. Since the two teams went a combined 16-8, you could certainly say it worked to some degree.

Pac-12

League correlation between size and S&P+: 0.007. Damn.

Once again, you can create a broad correlation here if you choose: the five biggest teams in the Pac-12 averaged a plus-9.8 S&P+, and the bottom seven teams were at minus-0.4. But obviously there are exceptions aplenty. Cal was by far the biggest team and did manage to go 7-5 despite a No. 64 S&P+ ranking, but Oregon State was the fifth-biggest team and was terrible, and Wazzu was easily the smallest team and went 10-2. Size is more of an identity than a predictor ... but it might still have some predictive powers.

AAC

League correlation between size and S&P+: 0.697.

I was sort of assuming that P5 conferences would follow the “size matters” rule pretty closely, and the ties would get looser at the lower levels. That certainly wasn’t the case with your favorite P6, the AAC, where three of the four biggest teams (UCF, Memphis, Cincinnati) were also three of the four best, and the two smallest were easily the two worst.

Still, there were exceptions. Temple and Houston succeeded with speed over size, and decent heft didn’t make good teams out of Tulsa or Navy.

That’s right, Navy: your fifth-biggest AAC team. Weird.

MWC

League correlation between size and S&P+: 0.313.

Here’s the ultimate “see what you want to see” conference. Among the five biggest teams in the Mountain West were a) easily the three best teams (Fresno State, Utah State, Boise State) and b) two of the three worst (UNLV, Colorado State). Granted, UNLV was distinctly decent before losing quarterback Armani Rogers to injury, and CSU was perfectly solid before this season. But still.

MAC

League correlation between size and S&P+: 0.672.

Miami’s Chuck Martin isn’t particularly good at maneuvering his way through close games — he’s got just about the worst record you’ll ever see in one-possession finishes — but he knows how to put some heft on the field. His receiving corps was enormous, and he fielded by far the biggest defensive back seven in the MAC. Miami was, on average, four pounds heavier per player than the second-biggest team in the league, Toledo.

In all, this conference lined up nicely. The five biggest teams had an average S&P+ rating of plus-3.8, and the seven smallest were at minus-13.1. And that sample of small teams includes conference champ NIU and WMU; the four smallest teams averaged a ghastly minus-17.4 S&P+ rating.

Conference USA

League correlation between size and S&P+: minus-0.212.

Well then. There’s always an exception. The five biggest teams in Conference USA had an average S&P+ rating of minus-11.3; the four smallest averaged minus-0.5 and included three of the five best teams in the league.

I will note, though, that while North Texas and UAB, two of the league’s three best teams, were pretty small overall, they were both large in the trenches. Marshall found success mostly through a tiny defense, so there are exceptions here, but UNT and UAB were both big where it seems to count the most. And MTSU, the C-USA East champ, was big in the secondary, which also seems to matter.

Sun Belt

League correlation between size and S&P+: 0.467.

This one lined up pretty well. The Sun Belt’s five biggest teams averaged a plus-1.7 S&P+ rating and featured only one bad team (USA), while the five smallest averaged minus-9.6 and featured only one particularly good one (option-heavy Georgia Southern). That said, App State was easily the smallest good team in all of FBS. The S&P+ top 30 averaged 238.2 pounds per player; your top-20 Mountaineers were 12 pounds lighter per play.

Indies

BYU’s quality hasn’t been where Cougar fans would prefer of late, but they certainly look like a power-conference team, size-wise, at least.

---

In the end, size ends up being the same type of general guideline as recruiting rankings: you are more likely to succeed when you’re big than when you’re small, but there are lots of exceptions in both directions.

• Average S&P+ rating for teams at 240 pounds or higher per player: plus-11.1
• Average S&P+ rating for teams between 235-240: plus-7.7
• Average S&P+ rating for teams between 230-235: minus-1.6
• Average S&P+ rating for teams between 225-230: minus-4.5
• Average S&P+ rating for teams under 225: minus-9.0

Indiana and Florida State were the only below-average BIG teams (240-plus), and MTSU and San Diego State were the only above-average SMALL ones (under 225).

Assuming I can figure out a decent way to isolate between size and recruiting ranking, then, this becomes a decent potential projection factor, yeah?