March Madness 

 As we often say, one of the goals of this blog is to share the conversations that take place around the halls of IQSS.  Well, the conversations at the Institute (along with just about every other office in the country) have been heavily slanted toward college basketball this week.  As I've posted here before, the relationship between sports and statistics has been both profitable for both sides.  And so, in that spirit, here are links to some recent papers on the NCAA Men's Basketball Tournament: 

  Identifying and Evaluating Contrarian Strategies for NCAA Tournament Pools  

 These authors (biostatisticians associated with the University of Minnesota) tackle one of the most important questions surrounding March Madness:  how do I maximize my chances of winning the office pool?  They find that, in pools that do not reward picking upset, strategies that maximize the expected score in the pool do not necessarily maximize the chances of winning the pool, since these brackets look too much like the brackets of other players.  Too late for this year, but maybe you'll get some pointers for next year.  For another paper that comes to a similar conclusion, take a look at  Optimal Strategies for Sports Betting Pools  . 

  
 


  March Madness is (NP-) Hard  

 Since it is too late to change your picks for this year, is there a way to tell when you don't need to pay attention anymore because you have no chance of winning?  A group of computer scientists from MIT consider this question, and show that the generic problem of determining whether a particular participant has been mathematically eliminated is NP-complete.  "Even if a participant were omnipotent in the sense that he could controll the outcome of any remaining games, he still would not be able to efficiently determine whether doing so would allow him to win the pool."  Of course, in a finite tournament with a finite number of players in the pool, it is possible to determine who could still win the pool.  I haven;t been eliminated yet, but things aren't looking too good.