Mar
23
Will Georgia Tech Software Predict NCAA Basketball Champ?
Three professors at Georgia Tech have developed a college basketball ranking system called LRMC that is designed to predict the outcome of the NCAA Division I basketball tournament. This software has been in use for a few years now, and it has been surprisingly accurate in predicting the winner of past tournaments. This year, the LRMC predicts the final game to be between North Carolina and Memphis, with North Carolina as the winner. Now that March Madness is upon us and there are only 16 teams left, I thought I’d check in on the LRMC rankings to see how the software’s predictions are holding up.
First, let me explain what the LRMC is and how it works. The software’s only input is basic scoreboard data from every Division I game played in the regular season this year. However, the Pure LRMC method takes into account not only which teams played and the final score, but the margin of victory and home court advantages to further boost its accuracy. The LRMC is particularly good at flushing out the best of the best and picking the true final four teams. It also excels at picking teams on the lower end of the rankings that will be the last (on the bubble) teams to be picked for the tournament. Taking the margin of victory and home court advantages into account helps the LRMC beat out most every other bracket predicting software out there, as shown in this chart.
So how is the LRMC doing this year? Their predicted winner for 2009 is North Carolina, which has been doing well in the tournament so far, and is still on track to win. The rest of its top five are Pittsburgh, Memphis, Connecticut and Duke, which have all made it to the Sweet 16. However, Memphis and Connecticut are both in the West bracket, so only one will make the Final 4; the same goes for Pittsburgh and Duke in the East bracket. Out of the LRMC’s top 16 teams before the tournament started, 10 teams are still remaining. Not too bad considering the LRMC does not take bracket match-ups into account at all; it only relies on data from the regular season.
Some of the missteps from the LRMC include Xavier (ranked 39th) and Arizona (ranked 36th) making it to the Sweet 16. The LRMC also ranked West Virginia as 10th overall, but they lost to 81st ranked Dayton in the first round!
This data shows how truly unpredictable college basketball can be. While the top teams are often identifiable, individual match-ups between middle-ranked teams are often no more than a coin toss. I recently read that the odds of successfully predicting the outcome of every game in the Division I tournament is over 1 in 9 quintillion. That is 9 thousand trillion! Those are way worse odds than winning the Powerball, which boasts an almost likely by comparison 1 in 195 million. So, I guess we can’t be too hard on the LRMC.
Last year the LRMC predicted that Kansas would beat Memphis in the final, which turned out to be true. It also picked out overrated teams that ended up losing in early rounds. Will the LRMC correctly predict the winner again this year with North Carolina? My bracket thinks so, but only time will tell.
Tweet
