Conference Realignment, Algorithmically

Using a computer algorithm to create the "perfect" 32 college basketball conferences

In the past few years, conference realignment has become a hot topic in the world of college sports once again. And a lot of the movements seem nonsensical. Texas and Oklahoma to the SEC? All these new Big 12 teams? UCLA and USC in the Big Ten? Come on.

Of course, there are reasons behind these moves. Teams and conferences are balancing media deals, contracts, and agreements to try and position themselves at the peak of the ever-more-competitive college athletics landscape, or just not fall too far behind (sorry, Pac-12). Geographic location and history matter less every year, as the influence of opportunities for massive media rights deals grows. And if you’re primarily a basketball fan like myself, you probably are further disheartened by the fact that realignment is almost exclusively driven by football, which is the cash cow of college sports.

But what if we could put all that behind us? Leave behind the mistakes of the past, the errors, the terrible moves, the nonsensical realignment, the corporate greed, the money, the capitalism — no, this isn’t a slogan for a communist revolution. We’re realigning the conferences, based on only what matters.

The Method, In Brief

We’re doing realignment based on only what matters — that is, if “what matters” to you is precisely these two factors:

• Geographic location

• Program strength (men’s basketball)

But if that describes you, then you’re going to love these new conferences. Let’s discuss the algorithm we’re using. Geographic location is trivial to quantify; we just use latitude and longitude.¹ For program strength, we’re going to use Kenpom’s program ratings. These ratings enncapsulate the strength of a program since 1997 in a single ranking, with some bias toward recency and stronger seasons. They are of course not perfect, but for our purposes they do a pretty good job of encapsulating the strength and desirability of a program in a single number. They also line up pretty well with what you’d expect. No points awarded for guessing the top 4.

Once we have these three data points for each school² (latitude, longitude, and program score), we simply plot these three-dimensionally, apply some adjustments, and separate them into clusters of nearby schools. If you’re a nerd like me and are interested in the exact algorithm for the clustering and some of the parameters I tweaked, I’ll describe those at the end of this writeup. But without further ado, let’s get into the new conferences!

Tier 1: The New Power 6

These six new conferences are pretty comfortably above the rest, forming the new “power” conferences. However, one of them is by far the strongest:

The Super League: Duke, Kansas, Kentucky, Louisville, Michigan State, North Carolina, Ohio State, Tennessee

This could arguably be Tier 0 by itself. Most of these schools are pretty large outliers in Kenpom’s program ratings, so they get put together without too much regard for geography. The others are fairly strong schools in the right geographical area to get pulled into this league instead of one of the other conferences.

The Big Eleven: Butler, Cincinnati, Illinois, Indiana, Iowa, Marquette, Michigan, Notre Dame, Purdue, Wisconsin, Xavier

All right, now this is a conference. Composed of many of the strongest midwestern programs from across the Big Ten, Big East, and others, this conference is incredibly strong, geographically compact, and features some great rivalries. Try and tell me you wouldn’t watch the hell out of this league.

The New Big East: UConn, Georgetown, Maryland, NC State, Pitt, Syracuse, Villanova, Virginia, Virginia Tech, Wake Forest, West Virginia

Despite a few new additions and notable omissions, this somehow feels more like the Big East than the current Big East does.

The Pac-12 But Cooler: Arizona, Cal, Gonzaga, Oregon, St Mary’s, San Diego State, Stanford, UCLA, USC, Washington

The strongest programs in the west (plus Cal and Stanford). What’s not to love?

The SEC?: Alabama, Auburn, Clemson, Florida, Florida State, Georgia Tech, LSU, Memphis, Miami (FL), Ole Miss, Mississippi State, Vanderbilt

A lot of underrated strong southern basketball programs here. Nice and regional compared to the sprawling current SEC.

Almost The Big 12: Arkansas, Baylor, Houston, Oklahoma, Oklahoma State, SMU, TCU, Texas, Texas A&M, Texas Tech

Another one that feels like a current conference but more compact. Should be a fantastic basketball league as well.

Tier 2: The Next Best

After the new Power 6, there’s definitely a noticable step down, but not a huge one. A lot of these teams are lower current power conference teams and higher mid-majors, creating some very competitive leagues.

The Central Conference: Creighton, Iowa State, Kansas State, Minnesota, Missouri, Missouri State, Nebraska, Northern Iowa, Tulsa, Wichita State

This conference is like the best parts of the MVC from a decade ago plus a few lower-end power conference teams that don’t quite fit in the new P6. Lots of fun basketball to be seen here.

The Northeastern Conference: Boston College, Penn State, Providence, Rhode Island, Rutgers, Saint Joseph’s, Seton Hall, St. John’s, Temple, VCU

Fun, local, northeast basketball with some great rivalries.

The New Mountain West: Arizona State, BYU, Colorado, Colorado State, New Mexico, New Mexico State, UNLV, Utah, Utah State, UTEP, Wyoming

Centered around the Four Corners states, this conference is more regionally consistent than the current MWC while still retaining a similar feel. I’d expect a lot of bubble teams just like the current conference.

The Appalachian Conference: Belmont, Davidson, Dayton, Georgia, Middle Tennessee, Murray State, Saint Louis, South Carolina, UAB, Western Kentucky

This conference has several of the most exciting mid-majors of the past decade or so. Pretty clearly the weakest of Tier 2, but also maybe the most fun.

Tier 3: The True Mid-Majors

This is it, the heart of college basketball. This tier has a lot of incredibly cool mid-major conferences that would be a thrilling fight for the autobid every year.

Return of the MAC: Akron, Buffalo, Duquesne, George Mason, George Washington, Kent State, Ohio, Old Dominion, Richmond, Saint Bonaventure

The center of this conference is shifted east a bit from the current MAC but it still retains much of its soul. Toledo is an unfortunate ommission though.

The Chicago Conference: Bradley, DePaul, Drake, Evansville, Green Bay, Illinois Chicago, Illinois State, Indiana State, Loyola Chicago, Milwaukee, Northwestern, Southern Illinois, Valparaiso

This conference has like five Chicago schools, I feel like it just has to be called this. As a Big Ten fan I’m a bit sad to see Northwestern fall to such a fate but let’s be real, their basketball history isn’t exactly far above a lot of these other schools.

The WCC 2: Fresno State, Long Beach State, Loyola Marymount, Nevada, Pacific, Pepperdine, San Diego, San Francisco, Santa Clara, UC Irvine, UC Santa Barbara

Only Nevada prevents this from being an all-Cali conference.

The Vine League: Boston University, Drexel, Harvard, Hofstra, Iona, La Salle, UMass, Northeastern, Penn, Princeton, Siena, Vermont, Yale

It’s like the Ivy League but just a tad less prestigious.

The Big North: Boise State, Eastern Washington, Idaho, Montana, Oregon State, Portland, Portland State, Seattle, Washington State

Hey OSU/WSU fans who are worrying about the Pac-12 collapsing and you getting left for the MWC; just look, it could always be worse.

The Small Fifteen: Ball State, Bowling Green, Central Michigan, Cleveland State, Detroit Mercy, Eastern Michigan, Marshall, Miami (OH), Northern Illinois, Northern Kentucky, Oakland, Purdue Fort Wayne, Toledo, Western Michigan, Wright State

Holy crap this one has a lot of teams. Formed from the less cool part of the MAC and the more cool part of the Horizon, this is a league that feels like it could be real.

The SoCon: Appalachian State, Charleston, Charlotte, East Carolina, East Tennessee State, Furman, Georgia State, UNC Greensboro, UNC Wilmington, Winthrop, Wofford

With how much the real SoCon has changed its teams you could tell me this is a real iteration of it from the past and I’d believe you.

The Fun Belt: Arkansas State, Little Rock, Louisiana, Louisiana Tech, North Texas, Oral Roberts, Rice, Sam Houston State, South Alabama, Southern Miss, Stephen F Austin, Tulane, UT Arlington

A solid group of mid-majors in the western South.

Tier 4: The Lower Mid-Majors

This is where things start to get rough. Unlike real lower mid-major conferences which tend to have some standout teams that can make some noise,a lot of these conferences have lost most of their best teams, leaving us with a whole lot of meh.

The Florida Conference: FIU, Florida Atlantic, Florida Gulf Coast, Georgia Southern, Jacksonville, Mercer, North Florida, South Florida, Troy, UCF

What a fall for UCF, from the Big 12 to this.

The New Big Sky: Air Force, Denver, Grand Canyon, Idaho State, Montana State, Northern Arizona, Northern Colorado, Southern Utah, Utah Valley, Weber State

Not much to comment on here.

The Patriotic League: American, Bucknell, Canisius, Cornell, Delaware, James Madison, Loyola MD, Mount Saint Mary’s, Navy, Niagara, Robert Morris, Towson, William & Mary, Youngstown State

The thought of Cornell and Navy in the same conference makes me giggle.

Literally Just The Summit League: Omaha, North Dakota, North Dakota State, South Dakota, South Dakota State, St Thomas, UMKC, Western Illinois

The only difference between this and the real Summit League is the lack of ORU and Denver.

The New York Conference: Albany, Brown, Colgate, Columbia, Fairfield, Fordham, Holy Cross, Lafayette, Lehigh, LIU, Manhattan, Marist, Monmouth, Quinnipiac, Rider, Saint Peter’s, Stony Brook, Wagner

If you think this conference is big now, you should see the agonizingly long period of time where I couldn’t get my program to stop spitting out 60-team conferences in this region. Stop being so close together.

The OVC, Kind Of: Austin Peay, Bellarmine, Chattanooga, Eastern Illinois, Eastern Kentucky, IUPUI, Jacksonville State, Lipscomb, Morehead State, Samford, Southeast Missouri State, Tennessee Martin, Tennessee State, Tennessee Tech

Can you tell I’m running out of things to say?

The Bigger West: Cal Baptist, Cal Poly, CSU Bakersfield, CSU Fullerton, CSU Northridge, Hawaii³, Sacramento State, San Jose State, UC Davis, UC Riverside, UC San Diego, Utah Tech

Another almost all-Cali conference, this time ruined by… Utah Tech? Why are you here?

The New South: Cambell, Coastal Carolina, Elon, Gardner-Webb, Hampton, High Point, Liberty, NC Central, Radford, UNC Asheville, VMI, Western Carolina

An interesting NC-centered conference.

Tier 5: The Low Majors

These are essentially equivalent to the current bottom-of-the-barrel conferences.

The Southernland: Jackson State, Lamar, Louisiana Monroe, McNeese State, New Orleans, Nicholls State, Northwestern State, Southeastern Louisiana, Southern

SWACtion 2 Electric Boogaloo: Abilene Christian, Houston Christian, Incarnate Word, Prairie View A&M, Tarleton State, Texas A&M Corpus Christi, Texas Southern, Texas State, UTRGV, UTSA

The NECxAmerica East: Army, Bryant, Central Connecticut, Dartmouth, Fairleigh Dickinson, Hartford, Maine, Merrimack, New Hampshire, Sacred Heart, St Francis (NY), UMass Lowell

And although there aren’t enough for me to justify separating out a new tier, the last 3 are a step below even the other low majors:

The SEAC: Bethune-Cookma, Charleston Southern, Florida A&M, Kennesaw State, NC A&T, Presbyterian, South Carolina State, Stetson, The Citadel, USC Upstate

Mostly The MEAC: Binghamton, Coppin State, Delaware State, Howard, Longwood, UMES, Morgan State, NJIT, Norfolk State, St Francis (PA), UMBC

SWinACtion: Alabama A&M, Alabama State, Alcorn State, Arkansas Pine Bluff, Central Arkansas, Chicago State, Grambling State, Mississippi Valley State, North Alabama, SIU Edwardsville

Hey, at least Chicago State found a conference.

Wrapping Up

I’m glad you joined me on this exploration of what CBB conferences could be in a more just world. Thanks for reading, this writeup is a bit of an experiment for me so I hope you enjoyed my shenanigans and terrible naming. If you’re interested in the nitty gritty of the algorithm I used and the code I wrote, I’ll be going in a bit more depth below. Otherwise, I hope you enjoy watching March Madness this year. Even if we have some mathematically non-ideal conferences in the real world, we’re always in for a treat in March.

The Deets

Now that the non-nerds are gone, I’m gonna let you in on a little secret: the algorithm I used here is actually super simple. Let’s go over it.

Here is the code I wrote. WARNING: BY CLICKING THIS LINK YOU CONSENT TO READING TERRIBLE CODE THAT I WROTE AT 3 AM. IT IS INEFFICIENT, UGLY, AND JUST GENERALLY BAD. DON’T SAY I DIDN’T WARN YOU.

First of all⁴, I preprocess the data. The kenpom program rating score needs to be multiplied by an adjustment factor to get the scale right, so that it’s weighted appropriately compared to geographic location. After a bit of tweaking I decided on a multiplier of 0.7. If you wanted team strength to be valued more you could increase this, and vice versa if you wanted geography valued more.

The algorithm I used is called k-means clustering. Despite all the scary math on the Wikipedia page, it’s very straightforward in practice. One drawback to this algorithm in real-world statistical usage is you need to pre-emptively decide on the number of clusters, which in many cases you might not know beforehand. In our case, though, I just picked 32 since that’s the current number of conferences, and it worked out fine.

The algorithm essentially works like this:

1. Assign the 32 conference centers to random points in the data range.⁵

2. Categorize every team by its closest conference center.

3. Move each conference center to the mean of the positions of the set of teams for which that conference center is the closest.

4. Repeat steps 2 and 3 a bunch. (I did 1000 iterations since it still ran pretty quickly but that might be overkill)

And that’s it! We get our conferences, just like magic.

Well… almost. This is what I was doing initially. The problem with this, however, is it doesn’t care about conference sizes, so I was ending up with a lot of 1-team and 60-team conferences. My solution was to multiply each distance by a penalty equal to [number of teams in the conference]^0.75. Why 0.75, you might ask? Again, mostly just trial and error. Too low and you still end up with too large of disparities in conference sizes. Too high and everything kind of just gets messed up.

The algorithm is technically non-determenistic because of the randomly chosen initial conference centers, but I didn’t “reroll” to get conferences I liked or anything. The result shown here was the only result I generated once I had the settings tweaked properly.

1. Due to the curvature of the Earth, technically changes in latitude will be a slightly greater ground distance than changes in longitude, and changes in longitude will be even less the further north you go. For my purposes, I’m ignoring this because I can’t be bothered to adjust for this and the effects should be pretty minimal.

2. The five new d1 schools in 2023 aren’t included in kenpom’s program ratings, so I have excluded them here too.

3. I actually also had to exclude Hawaii from my program because it was such an extreme lat/long outlier that it was messing everything up. I’m fairly sure this is where it would belong, though.

4. Technically the very first thing was getting all the data, which was a slightly tedious process of just going through this Wikipedia page and copying the GPS coordinates from each linked arena page. For a few schools where the arena didn’t have its own page, I used the school’s coordinates.

5. In this case I actually used the middle 60% of each axis of the data range, but this is mostly an artifact of when I was trying different things to fix an issue I was having earlier. In hindsight I don’t think this is necessary, although I don’t think it hurts either.