Jashon Robertson

SEC Schedule Proposal with FiveThirtyEight’s Model

Of the many interesting scheduling ideas floating around this off-season (including Joel’s here a few days ago), my favorite is from back in January: Nate Silver and FiveThirtyEight’s Make College Football Great Again. Their post uses the Big Ten, in part an effort to help prevent another Penn State or Ohio State playoff debate. We’re simply applying it here to the SEC. In terms of what is the most fair and the most fun, this is the best model I’ve seen.

No divisions, no conference championship game, nine regular season conference games.

First, ditching conference championship games frees up an additional week to be used on playoff expansion or an extra bye week.

If you’re going to go this route, you have to address the possibility of ties at the top of the standings. Models that include a handful of annual rivals and a rotation of other opponents – including ones that do it well like this one from SB Nation – either keep the conference championship game as a potential rematch, or shrug their shoulders at the notion of a tie.

In case you were born before divisions or overtime, ties are awful. No one is happy because literally no one wins.

I’m for anything that makes every single game matter as much as possible, and in this sense I don’t like playing conference championship games in non-divisional formats because of a higher probability of rematches. Divisional formats with annual rivalries greatly reduce the possibility of rematches: in 25 years the SEC has never had an annual rivalry (Tennessee-Alabama, Florida-LSU, Georgia-Auburn) play an encore in Atlanta.

But take out divisions? If you just sent the two best conference records to Atlanta, Tennessee and Florida would have run it back in 1993, 1995, 1996, 1998, and the very next week in 2001. This absolutely would have diminished the value of the regular season meeting.

Schedules need a way to preserve the rivalries that matter most, but also maximize the value of every game. And they need to be able to produce a champion that is rewarded in ways Penn State was not. Head-to-head needs to matter more, not less.

This is why I love FiveThirtyEight’s model:

Imagine a world in which historical rivals always play each other every year and yet, by almighty Rockne, the best teams in a conference always play one another, too. Imagine a world with no divisions.

Not only have I imagined such a world, my friends, but I have seen one. I have seen it in the hallways of a high-school debate tournament.

The solution that debate tournaments devised is something called power-pairing. Power-pairing just means that teams with the same record are paired off against each other, so that a team that starts off the tournament 2-0 will face off against another 2-0 team, for instance. It usually works by drawing the first two rounds of a tournament at random,1 and after that, everything is power-paired.

Three annual rivalries, two predetermined opponents, four flexed/power-paired match-ups

Here are the annual rivalries I went with:

  • Alabama:  Auburn, LSU, Tennessee
  • Arkansas:  LSU, Missouri, Texas A&M
  • Auburn:  Alabama, Georgia, Mississippi State
  • Florida: Georgia, South Carolina, Tennessee
  • Georgia:  Auburn, Florida, South Carolina
  • Kentucky: Mississippi State, Missouri, Vanderbilt
  • LSU:  Alabama, Arkansas, Ole Miss
  • Ole Miss:  LSU, Mississippi State, Texas A&M
  • Mississippi State:  Auburn, Kentucky, Ole Miss
  • Missouri:  Arkansas, Kentucky, Texas A&M
  • South Carolina:  Florida, Georgia, Vanderbilt
  • Tennessee:  Alabama, Florida, Vanderbilt
  • Texas A&M:  Arkansas, Ole Miss, Missouri
  • Vanderbilt: Kentucky, South Carolina, Tennessee

In FiveThirtyEight’s model, teams play their rivals in weeks 2, 4, and 7 of league play. Week 7 highlights the biggest rivalries as best we’re able. There’s a logistical point here for the SEC: this model works more smoothly for the Big Ten because they typically don’t start conference play until week four or five. So there’s a whole piece here about moving most or all non-conference games to early September that would have to be worked out, along with bye weeks.

In weeks 1 and 3 of league play, teams would face a predetermined opponent. FiveThirtyEight’s model uses the previous season’s standings to determine these foes:  Week 1 would feature teams from the top of the conference against teams from the bottom, Week 3 would feature best against best and worst against worst.

Opponents in weeks 5, 6, 8, and 9 of league play would be determined as the season played itself out. Weeks 5 and 6 would be decided after Week 4; Weeks 8 and 9 decided after Week 7, with both pairs of match-ups featuring one home and one away game. In each case, the league office would make the effort to power-pair teams based on their current records, creating the best available match-ups among teams yet to face each other. FiveThirtyEight’s piece had an algorithm help select these match-ups.

How would this look for the SEC? Here’s a sample season we played out (projected losses in red):

Let’s take Tennessee as an example. After four weeks the Vols are 3-1, and are paired with 4-0 Texas A&M and 2-2 LSU in Weeks 5 and 6. After seven weeks the Vols are 5-2, and are paired with 4-3 Ole Miss and 4-3 Auburn in the last two weeks of the season.

At the end of the year the Vols didn’t play Arkansas, Kentucky, Mississippi State, or Missouri. None of those teams finished above .500. Look at how many meaningful games everyone is playing:  Alabama, who wins the league at 8-1, played every team finishing 6-3 or better and three of the four teams finishing 5-4. South Carolina, who tied for last at 2-7, did not face Alabama or 7-2 Texas A&M.

If two teams tie at the top, power-pairing virtually assures they played each other during the regular season, thus head-to-head decides it. If three teams tie and their head-to-head results cancel out, power-pairing virtually assures you can find a next best common opponent to break the tie.

This would require some flexibility at all levels, especially with four unknown games to schedule on short notice. But of all the models I’ve seen, this one is the best at producing a worthy champion without divisions or rematches while sustaining key rivalries. It creates a sense of anticipation and opportunity as the season goes along with good teams continuing to face each other; in the above case the title is decided in the final week of the season when Alabama plays Texas A&M. It increases the value of every win: going 9-3 may not be cause for celebration right now, but against a schedule like this it becomes much more of an accomplishment (and hopefully creates healthier expectations along with more meaningful games). It also protects schools at the bottom of the league in any given year, helping struggling teams stay alive for bowl eligibility longer with more winnable games down the stretch. And, most importantly, it pushes opinion out of the equation and maximizes head-to-head results.

Leave a Reply

1 Comment on "SEC Schedule Proposal with FiveThirtyEight’s Model"

Notify of
Sort by:   newest | oldest | most voted
wpDiscuz