“Anyone can tell you that it’s going to be a close game,” write Bill Barnwell and Aaron Schatz, two of the smartest football analysts on the planet, in their Super Bowl preview at Football Outsiders.
The Green Bay Packers and Pittsburgh Steelers, indeed, are hard to distinguish statistically. The Steelers, counting the playoffs, have scored 24 points per game and allowed 15. The Packers’ numbers — 25 points scored per game, and 16 against — are nearly identical.
Bettors see two closely matched teams as well. Vegas lines have Green Bay favored, but only by a small margin — 2½ or 3 points, depending on the sports book. (Mr. Barnwell and Mr. Schatz also conclude that the Packers are slight favorites.)
Such close point spreads are fairly rare in the Super Bowl. Only 9 times in the 44 years of the game has the final point spread been within a field goal or less.
What’s interesting about those games is that while the point spread was close, the games themselves usually weren’t. The last one to feature such a tight point spread was Super Bowl XXXV in 2001, when the Baltimore Ravens, favored by just a field goal, instead blew out the Giants 34-7.
Prior to that was 1988, when the Washington Redskins — listed as narrow, 3-point underdogs — clobbered the Denver Broncos 42-10 in Super Bowl XXII. And before that was 1984, when the Los Angeles Raiders — also 3-point underdogs — steamrolled the Redskins 38-9.
Other games that looked as though they would be close, like Super Bowl XVI and Super Bowl V, in fact turned out to be. But on average, the final margin separating the teams in these games was 16 points.
Perhaps this is some sort of karmic retribution for the major upsets that the Super Bowl has seen over the years, like the Giants winning as 12-point underdogs over the Patriots in 2008. If Super Bowls that were supposed to be blowouts instead turned into close, epic games, it’s only fair that things turned ugly in a few games that looked like good matchups on paper.
In fact, the point spread — whether in the Super Bowl or in other games — is a much poorer predictor of the final score than most people realize. The chart below compares the closing point spread to the final score in all N.F.L. games played between 2002 and 2006.
The point spread was a relatively unbiased predictor of the final score: that is, teams favored by 5 points in fact won by about 5 points on average. The variance, however, was extremely high: on average, the point spread missed the final score by 10.4 points. The error was just as bad in games that were expected to be competitive as in those that looked like mismatches.
As a result, of the 487 games in the database in which the closing point spread was within a field goal, the game was decided by less than a touchdown (6 points or fewer) only 41 percent of the time. Games where one team won by double digits were somewhat more common, in fact, occurring 45 percent of the time.
This is not to suggest that there is no correlation at all between the closeness of the point spread and the closeness of the game. Of the games where one team was favored by 10 points or more, just 28 percent were decided by less than a touchdown. Still, the relationship is quite weak.
Here, then, is a bet that you might be able to sucker one of your friends into, especially after a couple of pregame beers. Offer to bet him that the game won’t be especially close: one team will win by at least 7 points. He may think the bet is too good to be true, but you have a 60 percent chance of winning.
