Data Analytics Could Win $1 Billion In March Madness Tournament

Metrifit Athlete Monitoring System

NCAA March MadnessWe admit it.  Here at Metrifit, we’re not only data junkies but we’re also huge college basketball fans.  With March Madness cranking up next week, we’re scouring the Interweb for any and all data analytics ideas that will give us an edge on our tournament brackets.

And with Warren Buffett throwing in a $1 billion (with a “b”) prize for the perfect bracket, any tips and tricks that we’ve learned from our work predicting athletic performance will be added to our slightly scientific solution for picking winners.  And if we didn’t already have a billion reasons to enter a bracket, Intel and Kaggle have also announced a new competition challenging groups of smart fans to create a model using machine learning and advanced statistics to predict the next national champion.

With a catchy name like March Machine Learning Mania, over 200 teams have already entered the first phase of Intel’s contest. Provided with stats from over twenty years of college games, their first task is to build a model that retrospectively predicts the winners of the previous five tournaments.  No prizes are awarded for this stage, but it does offer a warm-up validation that a team is on the right track.  Then, the real test starts on the first day of the 2014 tournament.

Except, Kaggle, a major player in “big data” prediction systems, wants to know more than just who the winner will be.  They want guesses on every possible match-up between all 64 teams, a total of 2,016 possibilities.  Also, just for fun, they also ask for a level of confidence in each pick.  It’s this combination of accuracy and probability that determines the winner. Saying Team A will be beat Team B with a 90% probability will hurt even more if Team B pulls off the upset.

As a sponsor, Intel is throwing in $15,000 to the team with the best record. Always on the lookout for analytics talent, Intel hopes to inspire data analysts and scientists with a fun contest.

“One of the greatest challenges we see is the lack of creativity in using data,” said Boyd Davis, Intel’s VP of the Data Center Group. “Intel’s goal with the competition is to drive more innovative uses for data. Kaggle is providing competitors with twenty years of historical game data broken down by teams, seasons, game by game results, game by game tournament results and seeds. With this type of information at their fingertips, competitors are poised to craft more accurate brackets – getting closer to that one-in-9.2 quintillion chance of predicting the tournament perfectly.”

About that 1 in 9.2 quintillion chance (that’s 18 zeros to the right of the decimal point), it’s a little hard to appreciate those odds.  To put it in perspective, David Sarno broke it down in his recent Slate article, “Another way to think about it: If all 317 million people in the U.S. filled out a bracket at random, you could run the contest for 290 million years, and there’d still be a 99 percent chance that no one had ever won.”

As a generous consolation prize, Quicken Loans, one of Buffett’s many companies, will award $100,000 to the 20 most accurate brackets.

As Metrifit knows, the future of sports performance belongs to those that can manage and embrace data.  “Sports teams are using new analytical capabilities to improve their team personnel and on-court performance,” said Davis. “It might be able to help coaches explore the connections between a variety of game conditions and the team’s performance, including how factors such as weather, time of day, travel schedules, team composition and the frequency of injuries could affect the likelihood of a win.”

So, finish up your next-generation, machine learning data model and get your brackets posted.  And if you need the most advanced athlete monitoring system, including trend analysis and prediction algorithms to manage your real team, please give us a call.