A Conference Realignment That Actually Makes Sense

Using K-Means clustering to create a reasonable college sports landscape

Adam Goldstick | Loyolan

I think I speak for the vast majority of college sports fans, especially those old enough to remember when regional rivalries actually mattered, when I say conference realignment has gotten out of hand. When Nebraska left the Big XII for the Big Ten in 2011 I, along with surely many others, were somewhat surprised and a bit disappointed by their abandonment of the classic rivalries they had with the likes of Colorado, Kansas, and Oklahoma.¹ However, I respected the freedom of schools to move to conferences which best suits their needs. But we now live in a world in which UCLA flies to New Jersey for a Big Ten tilt, Stanford competes in the Atlantic Coast Conference, and the Big XII spans from Central Florida to Provo, Utah.

While I fully recognize that money is behind the lions share of this realignment, is there a conference arrangement for that actually makes sense?

Data-Driven Conference Realignment

Using program-level data for all FBS college football teams and Division 1 college basketball teams, I ran a cluster analysis, via K-Means, to attempt to correct this conference realignment nonsense.² The statistics used were as follows…

These stats were normalized (placed on the same scale) via z-score normalization, and the schools were then categorized into one of five distinct regions: Northeast, Southeast, Southwest, West, or Midwest. I then ran a K-Means clustering algorithm within each region.

A New Football Landscape

This is the realignment, in regards to football, that resulted from this analysis.

Northeast Conference A: Boston College, Connecticut, Louisville, Maryland, Penn State, Pittsburgh, Syracuse, Virginia, Virginia Tech, West Virginia, Temple, Georgetown*, Villanova*

*-Non-Football members

Northeast Conference B: Akron, Army, Buffalo, James Madison, Kent State, Liberty, Marshall, Massachusetts, Navy, Old Dominion, Rutgers

Midwest Conference A: Cincinnati, Illinois, Indiana, Iowa, Kansas, Kansas State, Michigan, Michigan State, Missouri, Notre Dame, Ohio State, Purdue, Wisconsin

Midwest Conference B: Ball State, Bowling Green, Central Michigan, Eastern Michigan, Iowa State, Miami (OH), Minnesota, Nebraska, Northern Illinois, Northwestern, Ohio, Toledo, Western Michigan.

Southeast Conference A: Alabama, Auburn, Clemson, Florida, Florida State, Georgia, Georgia Tech, LSU, Miami (FL), Mississippi State, North Carolina State, Ole Miss, Tennessee

Southeast Conference B: Appalachian State, Charlotte, Coastal Carolina, East Carolina, Duke, Kentucky, Memphis, Middle Tennessee State, North Carolina, South Carolina, South Florida, UCF, Wake Forest, Western Kentucky

Southeast Conference C: Florida Atlantic, Florida International, Georgia Southern, Georgia State, Jacksonville State, Kennesaw State, Louisiana, Louisiana Tech, Louisiana-Monroe, South Alabama, Southern Mississippi, Troy, Tulane, UAB

Southwest Conference A: Arizona, Arizona State, Arkansas, Baylor, Colorado, Houston, Oklahoma, Oklahoma State, TCU, Texas, Texas A&M, Texas Tech

Southwest Conference B: Air Force, Arkansas State, Colorado State, New Mexico, New Mexico State, North Texas, Rice, Sam Houston, SMU, Texas State, Tulsa, UTEP, UTSA

West Conference A: Boise State, BYU, California, Oregon, Oregon State, UCLA, USC, Utah, Stanford, Washington, Gonzaga*, Saint Mary’s*

*-Non-Football members

West Conference B: Fresno State, Hawaii, Idaho³, Nevada, Nevada-Las Vegas, San Diego State, San Jose State, Utah State, Wyoming, Washington State

What Would This Change?

To see how this new landscape of college football would compare to what we have today, I analyzed the SRS scores (the Sports Reference team rating system) of the programs that would make up the new conferences, against the conferences of today.⁴

We see a slight decline in the range of conference SRS, indicating a narrowing of the gap between the best and worst conferences. There would be a decline in the variance in conference SRS, meaning greater parity between different conferences (i.e., less extreme differences in the average caliber of a conference). Additionally, a slight increase in the standard deviation of program SRS within conferences points toward a wider range of program quality within conferences. This results in greater talent dispersion within conferences, and when coupled with a smaller range in conference SRS, fewer "bad" or "dominant" conferences and an all-around more even playing field. On another note, schools would be much closer geographically. The average distance between teams within a conference would drop by nearly 322 miles!⁵

Average Distance Between Schools:

New realignment:

Today:

Rivalries Renewed

Needless to say, this would also have the benefit of reviving many regional rivalries of old, plus bring together nearby schools that have never shared a conference but have a history of hate.

• Pittsburgh vs. West Virginia (The Backyard Brawl)

• Oklahoma vs. Oklahoma State (Bedlam)

• Washington vs. Washington State (The Apple Cup)

• Georgia vs. Georgia Tech (Governor's Cup)

• Florida vs. Miami (FL)

• Clemson vs. Georgia

Just to name a few.

While the new conference landscape of today gives us big-time matchups week in and week out, there has been a definite shift in the balance of power. This sort of realignment I created is fun to fanaticize about, but unfortunately it will never happen as long as money is an object. For now, we can only dream…

Coming up…

Next week, I’ll publish an article about how this cluster analysis and imaginary realignment shakes out for basketball. Stay tuned!

1. Personally, I feel like Nebraska’s move was the first domino to fall, and kicked off the whole (albeit inevitable) realignment era.

2. An unsupervised algorithm that partitions data into ‘K’ clusters by minimizing within-cluster differences. It works iteratively by assigning points to the nearest centroid, updating centroids as cluster means, and repeating until convergence.

3. Historically was an FBS-level team, dropped to FCS in 20xx. Would return to FBS in this scenario.

4. Simple Rating System; a rating that takes into account average point differential and strength of schedule. The rating is denominated in points above/below average, where zero is average, although it should be noted that margin of victory has been assigned a lower bound of 7 points and an upper bound of 24 points. Non-major opponents are included as a single team in the ratings.

5. Talk about less travel costs!