Point Spreads, Again

I’m listening to “Mike and Mike” on ESPN radio, as I usually do in the morning, and they just spent the better part of five minutes talking about the point spread for the upcoming Super Bowl. The opening betting line has Pittsburgh favored by seven points, but some Las Vegas organization or another told them that the Cardinals would be underdogs to ten different NFL teams, had they made it to the game, including the Patriots and Cowboys, who didn’t even make the playoffs.

I have no idea who provided this information, or why they would even have betting lines for Super Bowls in alternate universes, but it’s worth reviewing the definition of the point spread again, because it’s an endless source of annoyance in these sorts of discussions:

The point spread is set at the point needed to get equal amounts of money bet on both teams.

In the ideal case, this is close to the actual score differential expert analysts would expect, but that’s not the definition, and there are fairly predictable ways in which the point spread for betting purposes will deviate from the expected point differential.

In particular, you would expect the point spread to be skewed a bit in favor of a team with a long history of winning and a strong nationwide fan base (say, the Steelers) that is playing against a team with less tradition and fewer fans (say, the Cardinals). There are more people out there who would be inclined to bet on the Steelers for non-football reasons than there are people who would lean toward betting on the Cardinals. As a result, you need to offer some enticement for betting on the Cardinals, in the form of the point spread.

The same basic argument works for the Cowboys and Patriots as well, in the case where you allow bets from parallel universes.

This is a generally known definition, and a widely recognized failure mode for the system, yet every year, we get a solid week of sports pundits talking Very Seriously about the point spread as if it were a real prediction of the outcome. Michael Wilbon on PTI is the only pundit I’ve seen regularly remind people of the difference between point spreads and predictions (as he did Monday)– everybody else treats them as if they were some sort of consensus expert pick. It drives me nuts.

Comments

  1. #1 ed kupfer
    January 21, 2009

    you would expect the point spread to be skewed a bit in favor of a team with a long history of winning and a strong nationwide fan base (say, the Steelers) that is playing against a team with less tradition and fewer fans (say, the Cardinals).

    And yet, point spreads routinely outperform regression-based forecasts based on objective predictors. The last time I tried this, it was for NBA games — I couldn’t find any systematic bias in point spreads that I could exploit. I don’t doubt that these biases exist, but I know they have very small effect sizes, so tiny they could probably be ignored.

  2. #2 bcooper
    January 21, 2009

    The argument that they represent a consensus expert pick arises out of something like the efficient market hypothesis. My understanding is that there are very few winning sports bettors, and while I’m sure that heavy juice by the sportsbooks is the primary reason for that, if the lines consistently had these sorts of systemic flaws you’d think there would be more.

  3. #3 melior
    January 23, 2009

    For a better definition of point spread, examine the interests of who sets them.

    Try this one: The point spread (like every line) is set at the point needed to generate the highest expected profit for the bookies.

The site is undergoing maintenance presently. Commenting has been disabled. Please check back later!