Analyzing March Madness — Completing Our Bracket Using Predictive Analytics

Every so often we get the opportunity to apply our knowledge to something a little different. It’s a great opportunity to break out and have a little fun with our data modeling. (For instance, our Halloween candy prediction last October.)

The Billion Dollar Bracket

In this case, we used our skills to complete a bracket for the Quicken Loans Billion Dollar Bracket Challenge, funded by Omaha’s own, Warren Buffet. Using past game data, and information about each team, we created unique models to predict the probability of each team’s chance of winning a particular match.
For our models, we took into consideration the significant variables that determine the typical likelihood of success for each team, things like assists, free throw averages, turnovers, and rebounds. We then applied these values to the respective teams for each matchup — letting us pit each team against each other more accurately.

Computers vs. People

We wanted to take the opportunity to apply one of our core beliefs.

“Computers are smarter than People. But a person plus a computer is smarter than the computer alone.”

We believe that a model is not complete without applying expert knowledge and experience. There is a certain amount of intuition that machines cannot take into account. Which is why when we produce models for our clients we use their data and the knowledge and expertise of their team.
To show this contrast, we let the math do the talking for our selections this year. We are especially interested in the matches that were predicted to be close calls, and upsets that the computer predicts with a high level of probability.

Without further ado, our bracket:

For the first round we predicted close calls in the following matches:

• Ohio State vs. Dayton – Ohio State likely to win (63.4%)
• Iowa State vs. N.C. Central – Iowa State likely to win (61.8%)
• New Mexico vs. Stanford – Stanford likely to win (59.8%)
• Memphis vs. George Washington – Memphis likely to win (55%)
• Cincinnati vs. Harvard – Harvard likely to win (55%)
• Gonzaga vs. Oklahoma St. – Oklahoma St. likely to win (52.4%)
• Duke vs. Mercer – Mercer likely to win (58.9%)

Be sure to check back after each round for an analysis of our predictions, and insight into why our predictions did (or didn’t) work. Please get in touch if you have any questions or want to learn more about using predictive analytics.