Just in case anyone is really good with probabilities but doesn’t know anything about American football betting, points spreads are a mechanism by which bookmakers equalize the amount of money gambled on a particular game. Rather than offer longer odds on an underdog, football games are bet according to a “spread”. You can’t bet on a game straight up (well, maybe you can, but it’s uncommon); instead, you get even odds betting on whether the underdog will lose by less than a certain number of points. Like odds, spreads move according to the amount of money invested on each side, so if the public is betting on the favorite the spread will increase (to get more people to bet on the underdog).

On sports talk radio and elsewhere, you tend to hear “experts” giving betting advice based on teams’ past performance against the spread, as in “the Steelers have covered the spread 9 times in their past 11 games.”

This seems like a really meaningless statistic. The whole point of the spread is to equalize the amount of money on each side of a game, so whether or not a team covers a spread should (in theory) be essentially random.

What am I missing?

Man am I a nerd; I thought this was going to be a question about impulse response functions.

It means that the odds makers are consistently underestimating that teams performance against opponents. Normally though when you look at ATS (against the spread) betting trends, you’ll see more specific information like if they’re more likely to beat the spread in home games for example or maybe in games against opponents in a certain division. In the example you gave, where a team has covered the spread in 9 of the past 11 games, you probably want to think twice about taking the underdog in an ATS bet (as compared to say a money line or over/under bet) in their next game. Although obviously if a team is going to cover the spread, they’re going to win a money line bet too.

Comparing points spreads, over/unders, etc have nothing to do with the game itself (football, hockey, soccer, etc) and everything to do with the handicappers ability to call the game.

If you’re aware that a bookie’s “predictions” as in "the Steelers have covered the spread 9 times in their past 11 games" have been consistantly high or consistantly low, that MIGHT give you an edge in placing your next bet.

I think the way the spread is set is irrelevant for this question. Consider it to be some unknown process, a black box which spits out a number; but it’s the same black box every week. A team’s performance against the spread is like looking at the recent history of a roulette wheel. Maybe the wheel’s out of balance, and certain numbers are coming up more often than pure chance would dictate. Maybe that football team has been playing against teams with a passionate fan base who have been betting out of emotion and shifting the spread away from a purely objective number. Or maybe it’s just random chance.

The spreads aren’t determined ultimately by bookies; they’re determined by the market with bookies reacting as bets come in. (Often big sportsbooks will let big customers help set the initial spread, perhaps offering them improved terms on early bets.)

Yes. According to this page, in recent years the Green Bay Packers, from a tiny town, beat the spread 70% of the time in division games, while “America’s Team”, Dallas Cowboys managed only 33%. (I think the figures are adjusted for pushes.)

Here’s a pdf paper describing some “feature search” regression method, though it didn’t seem to find anything as simple (and effective?) as Bet on the Pack; Bet against the Cowboys.

When using a bookmaker, you don’t get even odds: you have to risk more than you stand to win - typically, you risk \$11 to win \$10.

As RobotArm notes, bettors’ preconceptions - both rational and irrational - often affect their bets.

That all makes sense. Thanks, guys.