Gambling Strategies: Always Bet on White (Referees)

Justin Wolfers and Joseph Price showed us evidence last year that NBA referees tend to call less fouls on players of their own ethnicity -- and that this could influence the outcome of games.

Wolfers and Price (joined by Tim Larsen of BYU) have now turned their attention to the betting market's ability to exploit this inefficiency.

In research to appear in the Journal of Quantitative Analysis in Sports, the trio, I think unsurprisingly, find that point spreads don't appear to take into account the racial bias.

They write:

"This is a useful contrast to a long literature typically finding that betting markets are extremely efficient aggregators of information."

And wherever there is market inefficiency, there is profit opportunity. It turns out that a simple betting strategy of picking the team with the greatest racial similarity to the officiating crew when the crew is either all black or all white -- 4,493 games between 1991 and 2005 -- is successful in beating the spread 51.48% of the time. That's not quite profitable however, since you typically have to bet $11 to win $10 -- which requires at least a 52.38% success rate.

Through some statistical gymnastics, the researchers find that a betting strategy in which all referees are white and difference between the percentage of total playing time for black players on one team is at least 30% greater than for the other team yields a success rate of 56.76%. Unfortunately, there were only 508 games that had this configuration over 15 years. Still, the betting profit on this strategy would've been 7.48%.

A quick search of the playoffs finds one game that comes close to fitting the bill: Game 2 of the Raptors vs. Magic series. All three refs were white and the difference in playing time for black players for the Magic vs. the Raptors was 28 percent. But betting on the Magic wouldn't have been profitable here, as although they beat the Raptors 104-103, the Magic couldn't cover the 6.5-point spread.

(Hat tip: Sabermetric Research)

---