In a paper published in the Journal of Quantitative Analysis in Sports; Larsen, Price, and Wolfers demonstrate a profitable betting strategy based on the slight statistical advantage of teams whose racial composition matches that of the referees.

We find that in games where the majority of the officials are white, betting on the team expected to have more minutes played by white players always leads to more than a 50% chance of beating the spread. The probability of beating the spread increases as the racial gap between the two teams widens such that, in games with three white referees, a team whose fraction of minutes played by white players is more than 30 percentage points greater than their opponent will beat the spread 57% of the time.

The methodology of the paper leaves some lingering doubt however because the analysis is retrospective and only some of the tested strategies wind up being profitable.  A more convincing way to do a study like this is to first make a public announcement that you are doing a test and, using the method discussed in the comments here, secretly document what the test is.  Then implement the betting strategy and announce the results, revealing the secret announcement.