As I’m watching Villanova come from behind on a court that is only a couple of miles from their campus (indeed, it was reported that Villanova played on this court four times this season), I am curious to know if distance is a big factor in the tournament. It is widely know that seeds and locations are chosen in part to minimize the travel distance of the top seeds. Since seed and travel distance are not independent, I would expect a large negative correlation between travel distance and victory likelihood. I would like to know exactly how overwhelming this statistic is.
I found a short paper by GIS people that attempt to answer this question by comparing seeds, RPIs, and distance traveled (as well as differences between these three measures) to determine if distance makes a difference. Not surprisingly, they find that seed and RPI matter most. But seed and RPI are proxies for distance traveled, so this doesn’t really answer my question. Link.
If you have the answers to any of these questions, please answer them in the comments!
March 20th, 2009 at 7:54 am
I would imagine distance has an impact in defining a relative “home court” i.e. the closer to one’s campus the more likely that there will be fans of that particular school.
Yet I would like to add that proximity could be to generate more revenue for the NCAA tournament. There might be a correlation between higher seeds drawing more ticket sales. So if that were the case then maybe that proximity is more of a decision factor of a revenue model than it is of a winner outcome model.
March 20th, 2009 at 8:22 am
I know for the NBA that teams playing back to back games are usually more apt to lose than normal, which could be explained by fatigue from the previous night’s game or fatigue from travel. There could be something to distance. Might need to look just at lower seeds because the top seeds are usually placed close to home and are expected to win. Although, sometimes a high seed is “punished” and taken further away, i.e. the last team to get a 2 seed might get the 2 seed instead of a 3 seed but get placed further away. Sometimes its not punishment though, just scheduling problems that even if they would like to, they can’t have everyone go to somewhere close.
March 25th, 2009 at 8:44 pm
I want to disagree with the authors’ interpretation of their own results. Adding distance to a model that includes RPI and seed decreased AIC according to Table 1. This would seem to indicate that distance matters to the response controlling for the other “independent” variables. Just how much it matters might be discovered by comparing classification errors associated with the models.