KLBK Channel 13, the Lubbock CBS affiliate that is providing coverage of March Madness tournament action, came by my office today to do a story on statistical ways of looking at the games. The story was then broadcast on the late local news, following the buzzer of tonight's last game. I spoke about two topics. One was my own hot hand research, on streaks and other statistical oddities; KLBK was kind enough to plug the site, as shown above.

The other topic I addressed, which seemed to be the reporter's primary interest, was the use of statistical equations and the like to predict who would win particular games. I don't do this kind of research, but I was able to refer to the studies of Georgia Tech professor Joel Sokol, who does. In this article, Sokol and his colleague George Nemhauser show how their statistical tool (known as LRMC for Logistic Regression Markov Chain) has exhibited a better record of predicting NCAA tournament games than actual seedings, statistical ratings such as RPI and Sagarin, and media/coaches' polls such as AP and ESPN/USA Today. The LRMC may have worked well in the past, but this year it's laying a major egg.