By William Ezekowitz
There is perhaps nothing in American sports so spirited as a good old-fashioned college basketball rivalry. The home gym is packed with rabid fans, and the players are especially motivated to win one of the most important games on their schedule. But do these outside factors translate onto the court and make the results of rivalry games statistically different from non-rivalry games? Can we in fact “throw out the records when these two play”, as commentators often instruct us to do?
The metric I used to examine this was the betting line of each rivalry game. The thinking behind this was that betting lines are by their nature an efficient market, so once a spread is set, the underdog should cover the spread 50% of the time. If not, then there is an inefficiency somewhere in the methodology of creating the line. The advantage of using betting lines over something else (kenpom’s log5 formula, perhaps), is that the betting line measures popular consensus, and I was interested both in a possible statistical difference in rivalry games and in how that possible difference would manifest itself in comparison to popular thought. I examined 17 different rivalries, ranging from Duke/North Carolina to New Mexico/New Mexico State (unfortunately no one cared enough about the great Belmont/Lipscomb rivalry to make a line for it), and ended up with 139 games from the past eight years. The results were as follows:
In my sample, almost half of the time the underdog covered. Running a two-sample proportion T-test, I found that the result did not exceed an absolute value of 1.96, and therefore was not statistically different from the hypothesized value of 0.5. Though I did not start out trying to investigate this, I also found that in my sample home winning percentage, 69%, was almost identical to the median home winning percentage for college basketball teams last year. It makes sense that these two numbers would be similar, and adds further weight to the notion that the game being a rivalry has no eminent statistical effect on its outcome.
The conclusion that underdogs do not perform better in rivalry games is not necessarily the end of the story. It could be that underdogs do in fact perform better, and that superior performance is already factored into the betting lines. Moreover, it would be interesting to see how often home underdogs beat the spread, or even won the game. My sample contained only 42 instances of home underdogs, which is not a satisfactory number to answer that question. Statistically speaking though, at the end of the day, a rivalry game is just another game (don’t tell Duke fans).