College stats geeks' computerized system predicts Louisville to repeat as men's NCAA basketball champ

College stats geeks' computerized system predicts Louisville to repeat as men's NCAA basketball champ

The brainiacs behind Georgia Tech's LRMC computerized ranking system predict that the University of Louisville Cardinals (29-5) defend as NCAA men's basketball titlists, vindicating those upset that the team only has a #4 seed at the tournament's start this week.

The Logistic Regression/Markove Chain (LRMC) model comes up with its bracket selections based on scoreboard data collected during the regular season, factoring in stats such as margin of victory and looking at home vs. away results. The system is the product of four professors, three from Georgia Tech and one from Columbia University. Their areas of expertise include operations research, statistics, computer science and mathematics.

Their system has picked the ultimate winner three of the past six years, though last year incorrectly predicted Florida would beat Gonzaga for the championship. The LRMC did have eventual winner going to the Final Four though. The LRMC was last correct with its title pick in 2012, when the University of Kentucky won.

Once again, the Ga. Tech system did not have the pleasure of considering the school's Yellow Jackets team for the title. The team finished the season with a record of 16-17, not nearly good enough to make the tournament cutoff. (See the LRMC's bracket.)

Competition isn't only harder than ever for the Ga. Tech basketball team these days, as all sorts of stats models have emerged to predict results of sports events such as the March Madness tourney. The Prediction Machine, for example, is widely cited (and wants you to register if you want to find out its deep dark secrets about the NCAA tourney... though it too has Louisville going far.) FiveThirtyEight's Nate Silver he of the 49 of 50 states predicted correctly in the last U.S. Presidential election also puts Louisville as having the best odds to win this year, just slightly ahead of the University of Florida.

Read more about software in Network World's Software section.

Follow Us

Join the newsletter!


Sign up to gain exclusive access to email subscriptions, event invitations, competitions, giveaways, and much more.

Membership is free, and your security and privacy remain protected. View our privacy policy before signing up.

Error: Please check your email address.

Tags softwarebig dataapplicationsdata miningColumbia UniversityGeorgia Tech

Brand Post

Show Comments