The Historical | SEC Scheduling, Part IV

All statistics are courtesy of Football Outsiders, home of the F/+ Combined Ratings for college football.
The S&P+ rating was created by Bill Connelly; check out his college football analytics blog, Football Study Hall.

The previous articles in this series may be found in the “Articles of Interest” below.

What is all this nonsense again?

Last time I showed somewhat conclusively that, given how the team strengths in the conference were laid out^[1], since 2007 the cross-division scheduling had no statistically significant impact on the overall schedule strength, which as you’ll recall is a rather contentious topic around these parts. However, the analysis indicated the new 6-1-1 scheduling model instituted after Texas A&M and Missouri joined the conference in 2012 may be pushing the conference toward an unbalanced schedule — investigating that is what I’ll be doing today.

1 | Foreshadowing!

Projecting Team Strength

This is the trickiest part — how do you project the 2019 Vanderbilt Commodores or the 2021 Florida Gators when people “in the know” can’t even get it right on next year’s teams? On top of that, we all know that college football is cyclical, and although the SEC has been far and away the best football conference in America for about a decade now, the gap has been closed over the last few seasons, and the bottom has drop out eventually, right^[2]?

2 | Yes, I know, “FLAGGED”, etc.

I decided to go with the optimistic approach, however, and assumed that the SEC will remain the deep, talented league it’s been over the last eight years over the next eight. In that vein, I’ve used the 2007-2014 S&P+ numbers to project team strengths for 2015 – 2022. I did make a few adjustments to those numbers though, in order to better project where programs are heading into 2015.

I did nothing to the strengths of LSU, Georgia, South Carolina, Missouri, or Vanderbilt. The former four have had consistency in their coaching staffs (same head coach throughout the timeframe) and relatively consistent results on the field. There are few, if any, aberrant years, so there’s no reason to adjust their numbers. Vanderbilt, on the other hand, has had four different head coaches^[3] since 2007, which would suggest an adjustment might be appropriate moving forward. Fortunately, they are still Vanderbilt, and aside from the odd decent year (James Franklin’s first season, Jay Cutler’s senior year, etc.) they are awful on a pretty consistent basis.

3 | Bobby Johnson, Robbie Caldwell, James Franklin, Derek Mason.

I also did not adjust any of Arkansas’ numbers, for the same reason I didn’t touch Vanderbilt’s. Unlike Vanderbilt, Arkansas has been consistently decent, with the occasional great year (in this case, 2014, at least according to this metric), so I saw no reason to make an adjustment. For Alabama, I dropped 2007’s rating (3.1), because it was over 16 points lower than the next-lowest (2008’s 19.8) and is clearly not representative of where the program is under Nick Saban. Likewise I dropped Florida’s outstanding ratings from 2008 and 2009 (30.6 and 25, respectively); those are all-time great numbers achieved under Urban Meyer that are not a reasonable expectation of Jim McElwain’s tenure, at least not before he coaches a game in Gainesville.

For everyone else, I restricted the data to just the current coaching staff — that’s 2009-2014 for Mississippi State, 2012-2014 for Texas A&M and Ole Miss, and 2013-2014 for Kentucky and the Viles^[4]. I threw API a bone and included 2010 alongside 2013 and 2014, because I think we all know Malzahn was coaching the national championship team at this point.

4 | After the hate list piece last week, I determined this term more accurately describes my opinion of that school than Creamsicles.

With those adjustments, the averages look as follows:

Team	Average
Alabama	24.7
Ala. Poly	22.6
LSU	17.4
Texas A&M	16.6
Georgia	14.4
Ole Miss	14.3
Florida	13.7
S. Carolina	12.7
Missouri	12.6
Arkansas	12.2
The Viles	10.4
Miss. State	9.6
Kentucky	-1.0
Vanderbilt	-1.0

Given what I am about to do with these numbers, what this table shows you is where the teams are going to stack up, on average, over the next eight years given these adjustments. There are some questionable results here to be sure — API at #2 in the conference probably doesn’t hold, for one. I have a feeling Ole Miss is about to get nailed to the wall, and sanctions have a way of tanking your program for a couple of years. Texas A&M is trending downward, with their high placement buoyed by one outstanding year in 2012. I fully expect Arkansas to shoot up the chart over the next couple of years^[5], and regrettably things are looking up for the Viles as well. In general though, I think this is a fair place to start.

5 | Though it should be noted Wisconsin from 2007-2012 averaged a 12.3 in this metric, so who knows?

Using the averages above and the standard deviations of the adjusted data, I modeled each team in each season from 2015 – 2022 as a normally distributed variable. The idea here is to do a Monte Carlo simulation of the 6-1-1 method for the next eight years, which consists of modeling each team in each season as a random variable and running thousands of trials with a different set of team strengths each time to see what happens. What you put into the model is critically important, of course — in this case, it’s the average and standard deviation of each team’s distribution. Most teams’ standard deviation after the adjustments were similar to that before the adjustment, so I’m not too worried about that. The two exceptions were the two teams from the state of Alabama. In the Tide’s case I’m still using the bulk of the data, and dropping a clear outlier, so I’m still not worried. API is a bit more troubling, because they have been all over the place over the last seven years, whereas the three Malzahn years were all 20+ teams by this metric. Is Malzahn in that Saban/Meyer tier of coaches where he’ll produce a consensus top-5 team every year? Hopefully not, but for the purposes of this analysis I’m going to assume that he is on that tier and API will be that good every year.

Simulating the Schedule

Fortunately, we already know what the 6-1-1 schedule is going to look like, because the SEC has already told us — the cross-division rotation has been set through 2024. Using the same method I used in Part 2, I put in the schedule for all 14 teams, including home/road adjustments, and summed up the total SEC schedule strength and the total cross-division strength. As in Part 3, I found the coefficient of determination, and plugged it into the t-test to see if the relationship was significant or not. And then I automated that process and ran it 10,000 times.

In 9,874 of those trials, the relationship was significant. Put another way, 98.7% of the time the cross-division schedule had a significant impact on the overall schedule strength under the 6-1-1 scheduling system. Uh-oh.

Why is that? Well, look back at the table with the projected average team strengths. One potential explanation could be that Kentucky and Vanderbilt are terrible in this simulation, whereas Alabama and API are dominant. That means four other teams^[6] are getting wildly different permanent cross-division games compared to the rest of the conference, which all fall into a relatively tight band. This is the gist behind the LSU argument against the system, although it doesn’t apply to them at all in this simulation. So how do we fix this?

6 | Mississippi State, Ole Miss, the Viles, and Georgia, respectively.

The 6-0-2 Model: No More Permanent Rivalries

Let me be absolutely clear here: I don’t want this, and I don’t want to see permanent rivalries go by the wayside. I want to see API-Georgia, Bama-Viles, and LSU-Florida happen every season. I think the new Arkansas-Missouri series is going to be great. But, it’s worth determining if relegating those games to a less frequent interval makes the schedule more “fair”.

To do this, I had to determine a new rotation for the cross-division games, which I seeded with the current rotation for the 6-1-1 model as provided by the SEC. I elected to take a page out of the pre-1992 scheduling playbook, and set up a system based on rotating home-and-homes. You play a team at home one season, you play at their house the next season, and then you don’t see them again for a few years. Just as an example, Alabama’s rotation would look like this (home team in bold):

Year	Opponents
2015	Georgia	S. Carolina
2016	Florida	Georgia
2017	Viles	Florida
2018	Kentucky	Viles
2019	Vanderbilt	Kentucky
2020	Missouri	Vanderbilt
2021	S. Carolina	Missouri
2022	Georgia	S. Carolina

Running through the same process as for the 6-1-1 model, the relationship was significant in 9935 trials, which is… worse, and although there is probably a set of 10,000 trials^[7] that is “better” with this system than that with the 6-1-1 system, effectively that’s the same result.

7 | Excel calculates random numbers to 15 decimal places, and each trial involves 112 random team strengths, so that’s 112 quintillion possibilities.
Not all of those are unique, and many are effectively identical, but you get the idea.

What about a 6-0-2 schedule without rotating home-and-homes, where a team plays at least one game against every team in the other division every four years? This is a departure from the current rotation (done for convenience), but that would look something like this for Alabama:

Year	Opponents
2015	Florida	Georgia
2016	Kentucky	Missouri
2017	S. Carolina	Tennessee
2018	Vanderbilt	Florida
2019	Georgia	Kentucky
2020	Missouri	S. Carolina
2021	Tennessee	Vanderbilt
2022	Florida	Georgia

Using the same methodology above, the relationship between cross-division schedule strength and SEC schedule strength was significant in 9914 trials, so no change from before. At this stage it’s pretty apparent what the issue is, but humor me a little longer — what if we added another game to the schedule?

The 6-1-2 Model: An Additional Rotational Opponent

This would be the model everyone was clamoring for when Texas A&M and Missouri joined the conference — let’s not mess with the cross-divisional schedule whatsoever, and just add a ninth game to account for the additional in-division opponent. Certainly with the move to a college football playoff and a corresponding emphasis on schedule strength^[8], yet another SEC game would send the SEC champ’s schedule strength even higher than it has been over the last few years.

8 | Hi Baylor!

Of course, nine is an odd number^[9], which means half the league’s getting an extra home game compared to the other half under any nine game system. I suspect that’s the main reason we won’t be seeing a nine game schedule anytime soon in the SEC, but for the purposes of discussion let’s just assume it’s on the table. Once again, Alabama’s rotation would look something like this:

9 | This kind of deep insight is why you keep coming back, isn’t it?

Year	Permanent	Rotational
2015	Viles	Georgia	Florida
2016	Viles	Kentucky	Georgia
2017	Viles	Vanderbilt	Kentucky
2018	Viles	Missouri	Vanderbilt
2019	Viles	S. Carolina	Missouri
2020	Viles	Georgia	S. Carolina
2021	Viles	Florida	Georgia
2022	Viles	Vanderbilt	Florida

You’ll note I went back to a rotation based on back-to-back home-and-homes here, since that didn’t seem to make much of a difference with the eight game schedule. Running the simulation again, and this time it’s 9916 trials in which the correlation was significant. I also did this for a 6-2-1 schedule, going back to the two permanent rival system of the post-1992 SEC, and got a similar result. Just for fun, I set up a 16 team league (bringing in North Carolina State and Virginia Tech) with four divisions and one permanent cross-division rival for each division (a 3-3-3 schedule)… and got the same result. Not sure how you’d go about determining the champion without a mini-playoff, but that would be a fun league^[10]!

10 | Not going to get into the details of that setup now, but if you want to know more let me know in the comments.

So, what’s the problem here?

This is puzzling, right? The schedule seemed to be fair before, but regardless of how you set it up it appears as though it’s going to be unbalanced moving forward. To see why, let’s take another look at those average strengths again, but this time broken out by division (overall conference rank in bold):

Team	Average
Alabama	24.7 (1)
Ala. Poly	22.6 (2)
LSU	17.4 (3)
Texas A&M	16.6 (4)
Ole Miss	14.3 (6)
Arkansas	12.2 (10)
Miss. State	9.6 (12)
Georgia	14.4 (5)
Florida	13.7 (7)
S. Carolina	12.7 (8)
Missouri	12.6 (9)
Viles	10.4 (11)
Kentucky	-1.0 (13)
Vanderbilt	-1.0 (14)

And I think the issue becomes clear. As long as the East continues to lag behind the West in terms of overall strength, no cross-division rotation is going to be fair — the East will always have the tougher slate, because the West has stronger teams. The last six conference champs have come from the West for a reason, folks. The East dominated the conference championship game for the first decade and a half, highlighted by six straight victories from 1993 – 1998, but it’s been all West in the advanced metrics era, and frankly I don’t see that ending anytime soon, barring a significant realignment of divisional structure. Keep in mind Spurrier is retiring any day now, which could mean South Carolina returns to what they were in the 1990s, which was not very good. Nobody knows for sure what McElwain will do with Florida, and Georgia still hasn’t figured out they can only go so far with Mark Richt at the helm. Kentucky and Vanderbilt will always be Kentucky and Vanderbilt.

Meanwhile, API is unfortunately ascendant, LSU and Alabama show no signs of slowing down, Arkansas is improving greatly under Bret Bielema, and if Ole Miss can shake the NCAA they may be on to something in Oxford. So, instead of complaining about the schedule and how such and such team has to play such and such team every year, complain about the East not living up to their end of the deal, because that’s the real problem here^[11].

11 | #hottake alert!

ROLL TIDE