NCAA Basketball: Can advanced stats predict NCAA tournament success?

How useful are advanced stats when predicting NCAA Basketball Tournament championship outcomes? 

The most important thing about March Madness for some people is being able to win their bracket pool. There’s nothing better than showing that annoying person at the office that you know more about college basketball than they do. Everyone rips themselves apart over these tournament match-ups in hopes of securing bragging rights until next March.

But let’s be honest with ourselves: so much of filling out a bracket and winning a bracket pool is about… 80 percent luck. Did it really take any knowledge for people to predict UAB would beat Iowa State last year, or was it just a stab in the dark that happened to be right? Can we really fault anyone for putting Iowa State in the Sweet 16 last year? Probably not.

Predicting single games is nearly impossible. In sports, particularly college basketball, cliches like “any given Sunday” really do exist.

Sure, a 16-seed has never beaten a one-seed, but a handful of 15-seeds have knocked off two-seeds, including three in the past four years alone. You can’t help but feel that a 16-seed is going to get a victory eventually. 

However, while predicting one, single game may be extremely difficult, is there a method towards predicting who will come out on top?

I dug into some stats. What do most of the past 10 NCAA champions all have in common? Two main things:

  1. They were ranked in either the top 10 of kenpom.com’s adjusted defensive efficiency, or adjusted offensive efficiency (the only team that wasn’t was 2011 UConn, who finished 13th in defensive efficiency, and 18th in offensive efficiency).
  2. They finished in the top 10 of kenpom.com’s final rankings, which is calculated using an algorithm that incorporates scoring margin, strength of schedule, offensive efficiency, defensive efficiency, etc (all 10 past champions did this).

Out of curiosity, I plotted the past 10 champions by their adjusted offensive efficiency, and their adjusted defensive efficiency. Here’s what it looks like (click here for a bigger version):

Graph of former championsIdeally, teams want to find themselves on the bottom right corner of this graph. That means they were great both defensively and offensively. 2008 Kansas is the only team that finds themselves firmly in that corner, with 2012 Kentucky, 2010 Duke, and 2013 Louisville being close.

In the upper right corner, teams like 2015 Duke, 2007 Florida, and 2009 North Carolina were fantastic offensively, but not quite as great defensively, but still great, obviously, because they were, indeed, national champs. 2006 Florida is right in the middle.

The top left corner is bad. Maybe Connecticut got kind of lucky the two years they won. But they did have guards fully capable of putting the team on their back.

So what teams fit these two criteria this year? Just eight teams: Kansas, Virginia, West Virginia, Villanova, Kentucky, Michigan State, North Carolina, and Oregon. I graphed them on a similar plot, and that looks like this (click here for a bigger version):

2016 Contenders GraphAgain, ideally, a team wants to find itself in the bottom right corner. That means they are great both offensively and defensively, which is important. While my personal criteria only required teams to be in the top 10 of either adjusted offensive efficiency or adjusted defensive efficiency, every team except for 2014 UConn found themselves in the top 25 of both categories.

So statistically, Kansas, Virginia, North Carolina, and Michigan State are the best picks to win it all in your bracket this year.

But notice the difference in the range between the two graphs. The range in adjusted defensive efficiency for the past champions is 85.5 to 94, while on this year’s contenders it’s 91 to 100. This speaks to the down year that college basketball is having this year.

If we were to plot this years contenders on the same graph with the same domain and range as the champions, only Kansas, West Virginia, North Carolina, Virginia and Villanova even show up (click here for a bigger version):Graph Former + ChampsPretty interesting. Notice the lowest adjusted defensive rating of the 10 past champions is 2009 North Carolina, who was at 92.9. Only three teams this year have an adjusted defensive rating lower than that, and looking at the graph, that would be West Virginia, Virginia, and Kansas.

What does all of this tell us? Quite honestly, it’s hard to tell. Historical precedent tells us that really only eight teams have a legitimate chance to hoist the trophy in April. But if we apply an even stricter historical precedent, it’s limited to just three teams.

Still, even if we limit it to Kansas, Virginia, West Virginia, Villanova, Kentucky, Michigan State, Oregon, and North Carolina, does that mean no other teams even have a chance? What about Oklahoma and Xavier, the two highest seeds that don’t fit the criteria?

Keep in mind that both Connecticut teams were major outliers in this. The 2014 team was 10th in adjusted defensive efficiency, and 39th in adjusted offensive efficiency.

The 2011 team was 13th in adjusted offensive efficiency and 18th in adjusted defensive efficiency. So it’s not impossible for another team to crash this party. Maybe we get another outlier this year? That certainly makes it seem like a much bigger possibility for Oklahoma, Oregon, or Miami to win it all.

We won’t know for sure until a couple of weeks. But if you want to secure bragging rights in your office, or among your group of friends, historical precedent says you should pick one of the eight teams listed above.

If you want to go off of an even stricter basis, pick one of Kansas, West Virginia, Virginia, North Carolina, or Villanova. History does tend to repeat itself, and you’ll have a much better chance if you pick one of these teams.

Or don’t. As always, it’s up to you. Let the madness begin.