Using Statistical Data and Analytics to Accurately Predict March Madness
Most of my life I was a nerd in disguise. In high school, my letter jacket covered with wrestling state championships and football league honors was my disguise. It wasn’t until after college when I completely became accepting of my nerdy-ness.
I was often successful in sports not because of my superior athletic ability but because I outsmarted my opponents. I always knew the score or what I needed to do, and I wasn’t going to beat myself. Of course, my ability to remember scores of matches and devotion to statistics helped too. This ability also caused many of my coaches, teammates, and close friends even to start calling me ‘stat man’ in high school. Still, the average person in high school just saw me as a jock.
Despite my math abilities and passion for sports, I never applied any statistical analytics to the biggest prediction event in sports, the NCAA Basketball Bracket Challenge. Traditionally, I do average to below average in bracket pools. Every year, it seems that the more basketball I watch that year, the worse I do in my pool.
However, two weeks ago I saw a video on ESPN talking about how every team that has won the NCAA tournament in recent years had consistently been towards the top of offensive efficiency, defensive efficiency, RPI (Rating Percentage Index), and offensive rebounding. I had some down time on the Monday after Selection Sunday, so I decided to develop a formula using these metrics. I would also add conference RPI and Tournament seed.
I completed a total of twelve brackets. The first one was me as a fan guessing, which I used as a control. In the other eleven, I used the different metrics. On some brackets I excluded the metrics for seeds, others I excluded wins, others I excluded conference RPI, etcetera.
So how did my brackets look?
Every bracket had either Gonzaga, Villanova, or North Carolina winning them. Not surprising considering they were all number one seeds in the tournament. To the rights is my first bracket that took into effect RPI, Wins, Offensive Efficiency, Defensive Efficiency, and Offensive Rebounds. Utilizing these metrics was the first bracket I completed and coincidentally the best one.
How did I do overall?
As I said earlier, I entered a total of 12 brackets. Nine, including my control, were done on ESPN.com. I completed three on BracketChallenge.NCAA.com. All of my brackets on ESPN.com, except my control, are in the 92nd and 99th percentiles on ESPN, currently. The two worst are the ones with Villanova winning the tournament. Of the three brackets on NCAA.com, They are between the 92nd and 100th percentiles. All of my brackets correctly predicted North Carolina and Gonzaga, and the bracket I titled ‘Zagathon,’ which was the first bracket I filled out utilizing RPI, Wins, Offensive Efficiency, Defensive Efficiency, and Rebound Rate has three of the final four teams selected.
Where were my numbers were off?
For those of you paying attention to the tournament, you might wonder about South Carolina, the seven seed who made the Final Four. Where did I have them? Don’t tell Frank Martin, but my numbers had them 43rd of the 68 teams. In every bracket I had them losing to Duke who they would end up defeating 88-81.
In fact, the entire East Regional was a mess in all of my brackets. In every single one, I had Villanova getting to the final four, and in many brackets I had them defeating SMU to get there. Both SMU and Villanova were the most underperforming teams in all of my brackets.
Xavier, who just lost to Gonzaga this weekend in the Elite Eight, was another place where my numbers were way off. While a few of my brackets had them upsetting Maryland, not one had them getting past the second round.
What to take from this?
So March Madness is fun and everything, but why am I talking about NCAA basketball numbers on a marketing consulting blog? Once again, I have used my math and nerdy-ness to make me look good at sports. I’ve even already clinched my pool championship with some of my former Loyola MBA classmates. This way of utilizing analytics to predict is a very similar to how Emblem and I generate and utilized data to design and enhance your marketing strategically.
Most of us are experts in our relative fields. We spend 40+ hours per week focusing in on all aspects of our company and industries. Even then, we don’t see everything. While I am a casual fan of college basketball, I never played basketball, and I am certainly not an expert. However, by simply utilizing statistics and analytics, I was able to produce eleven brackets, which as of now, have all outperformed every ESPN expert. This group includes former basketball players and coaches who eat, sleep, and breathe college basketball. Not much different than how well you know your own business and industry.
Sure, consumers and markets can be as unpredictable like South Carolina and Villanova’s basketball teams. However, ultimately data can be an excellent predictor of how we should strategically market our companies’ services and products. Without it, we are only guessing.