The NCAA men’s basketball tournament is considered one of the leading wastes of time in American business, so it should come as no surprise that your friendly blogger would search high and low to find a way to mix his love of college hoops with a little thing his bosses like to call “work.”
It seems a trio of college professors has been employing predictive analytics to predict the at-large teams that make up the NCAA tournament for the last 20 years. Using the analytics software from SAS, the professors have enjoyed a 94 percent success rate in putting together what they call their Dance Card. Two years ago, the professors were correct on 33 of the 34 at large bids (the rest of the field is made up of teams that earn guaranteed bids through their league championship tournaments.
“SAS analyzes huge amounts of information, pulls out what's important and ignores what's not. SAS predicts what's coming in the future, as opposed to simply describing what happened in the past," says Jay Coleman, an operations management professor at theUniversityofNorth Florida.
Coleman explains the Dance Card reflects the trend in analytics toward more prediction and less description.
"Predictability is a key part of business analytics that helps companies move from information to a decision, from the descriptive to the predictive," says Coleman. "With forecasting, data mining, optimization and other advanced predictive software from SAS, businesses can go beyond simply reporting on what has already happened and instead understand what will happen and where to go next."
So how did the profs do this year? About average. For 2012, the NCAA men’s tournament has 31 automatic bids and 37 at-large bids. The Dance Card correctly picked 35 of the 37 for a 95 percent success rate.
The analytics predicted theUniversityofWashingtonandDrexelUniversitywould make the field. Instead, the selection committee choseColoradoStateand theUniversityofSouthern Florida.ColoradoStatewas the team closest to the Dance Card bubble, but USF jumped over seven other schools to earn their spot in the tournament.
We’ve been writing a lot about predictive analytics lately and we’re helping our friends at Robert E. Nolan Co. distribute a survey on how insurers are using these tools. The survey results will be shared in a Webinar on March 28, and in a whitepaper, which will be presented after the Webinar.
In the meantime, I need to get my hands on some good models which will help me predict the winners in the tournament when the games start this week so I can make a little money on this. Anyone who has rolled out some analytics on who will win all the games is advised to send their predictions my way.
Be warned, though. I will judge the quality of your picks by whether or not you predict my Xavier Musketeers will defeat Notre Dame in their game on Friday night. After all, emotion usually overrules cold hard facts. Isn’t that why the house always wins?