≡ Menu

Why is the Over/Under So Low In Super Bowl XLVIII?

The Over/Under for Super Bowl XLVIII was just 48 points for most of the last week, although it went up to 48.5 on Thursday and may be at 49 by kickoff. In any event, such a low number should strike Broncos fans as really odd, since the average over/under in the team’s first 18 games was 53.4 points. And the “Over” has hit in 10 of those games!

Seattle, meanwhile, has had an average over/under of 42.8 points. As it turns out, the “Under” has hit in 12 of Seattle’s 18 games this season, including each of the last seven. Readers who are good at arithmetic might have already noticed that the average of 53.4 and 42.8 is 48.1 points.

The graph below shows the Over/Under in each game this season for Denver and Seattle:

As you can see, 48 is a very high O/U for Seattle and a very low O/U for Denver. Intuitively, it makes a lot of sense that the over/under would represent the midway mark for both teams. In reality, that’s not the case. On average, the Super Bowl has an over/under that is nearly six points higher than you would expect. Allow me to walk you through the table below. In 2012, Baltimore’s average Over/Under in all games that season other than the Super Bowl, was 45.5, while San Francisco’s over/under was, on average, 42.7 points. An average of those two numbers is 44.1, which we could call the “expected” Super Bowl over/under. In reality, the Super Bowl Over/Under last year was 47, which means the actual line was 2.9 points higher than expected.

YearWinnerAvg O/ULoserAvg O/UAvgSB O/UDiff

From 1990 to 2012, the actual Over/Under in the Super Bowl was higher than we would have expected. And, on average, there was about six points of “inflation” in the Super Bowl Over/Under. But there are zero points of inflation in this year’s game. Which brings us to the question of the day: Why is the Over/Under so low?

Note that this year’s line isn’t a reaction to historic mispricing: the “Over” hit on 24 of the first 47 Super Bowls, which includes a 9-9 split in the last 18 games. The story in previous years has been “set a high Over/Under, watch half of the games go over, anyway.”

But if anything, the Over/Under is even lower than you might think. I ran a regression to predict the Over/Under of each Super Bowl from 1990 to 2012. As my inputs, I used four variables: the points per game scored and allowed (including playoffs) of each team entering that game. The best-fit formula was:

-3.8 + 0.74 * TmA_PPG + 0.36 * TmA_PPG_Allowed + 0.74 * TmB_PPG + 0.36 * TmB_PPG_Allowed

This year, Denver has scored 36.4 points per game and allowed 24.0 (including playoffs), while Seattle has averaged 25.7 PPG and allowed 14.6. If you put those numbers into the formula, we would expect an Over/Under of 56.1 points. That means the Over/Under in Super Bowl XLVIII is 8 points lower than we would expect, given these two teams.

So how do we explain this year’s Over/Under?

1) Weather. One could argue that the first cold-weather Super Bowl might have caused Vegas to go ultra-conservative on the Super Bowl. But the numbers don’t bear that out: some preliminary research I conducted suggests that a cold-weather game lowers the over/under by somewhere between a half-point and a point, not six-to-eight points. And there is no research to indicate that cold-weather games produce significantly less scoring.

2) Recency Bias. Seven straight Seattle games have gone under, along with five straight Broncos games. That’s a pretty significant trend, and perhaps — along with the slump Russell Wilson has been in — an explanation for the low number.

Vegas thinks people want to bet against this guy

Vegas thinks people want to bet against this guy.

3) The Books Believe in Seattle. Perhaps the most nefarious of reasons is the one that intrigues me the most. The numbers seem to point in the direction of Seattle. Advanced NFL Stats and Football Outsiders both grade the Seahawks as the better team, and Bill Barnwell and several others (including me) picked Seattle to win. That anecdotal evidence doesn’t mean much, but consider other pieces of evidence:

  • On January 16th — i.e., after the division round of the playoffs — Bovada had the NFC as a 2.5-point favorite in the Super Bowl. At the time, Bovada listed Denver/Seattle as the most likely Super Bowl matchup, offering 13/10 odds.
  • A week earlier, with 8 teams remaining, Denver/Seattle was still the favorite (9/4), and the NFC was still a 2.5-point favorite.
  • Entering the final week of the season, Bovada listed Denver and Seattle as its two most likely Super Bowl teams, and had NFC as a 2.5-point favorite.
  • After defeating San Francisco, Seattle opened as narrow favorites against the Broncos. Bovada, after sticking with a line of 2.5 for awhile, opened with Seattle as a 1-point favorite.

Now, though, everyone has listed Denver as a 2-3 point favorite. So what gives? The Super Bowl is a unique gambling event. The action placed on the Super Bowl dwarfs what Vegas will sees for every other game, making it crucial that Vegas gets about half as much action on both sides. Even if Vegas thinks Seattle is the better team, no book wants to keep Seattle as a favorite and invite every better who wants to put money on Peyton Manning to come to their book. That’s a recipe for bankruptcy or striking it rich, an undesirable outcome when the book can play it straight and collect its 5%.

But, if the books do believe in Seattle, one way to cash in on an undervalued Seahawks team is to set an artificially low Over/Under, thereby inviting betters to place money on the Over. I suspect most of us believe the odds of a Seattle victory and the game going Under are strongly correlated; if that’s the case — and I grant that reasonable people can differ on this take, but I believe it is — then setting a low Over/Under is Vegas’ way of saying they believe in Seattle, but don’t want to expose themselves to catastrophe by making the Seahawks the favorites. If Vegas likes Seattle, a low Over/Under (by Super Bowl standards) is a much safer alternative than moving the Seahawks the favorite.

4) Pace. There is a fourth explanation, and it is more palatable to the eyes of Broncos fans. During the regular season, the Broncos had 192 drives, the 9th most in the league. That’s because Denver played at a very fast pace, and incorporated a lot of no huddle: according to Football Outsiders, the Broncos had the 3rd fastest pace in the league. Seattle, meanwhile, ranked in the bottom eight of the league in both drives and pace.

But in the playoffs, Denver has shifted philosophy. Against the methodical Chargers, Denver employed a low-tempo offense, and wound up having just eight drives all game. Against New England, Denver again saw just eight possessions. The Broncos had 71 drives for touchdowns during the regular season, and none of them lasted seven minutes. In the playoffs, the Broncos have had three touchdown drives that lasted 7+ minutes.

Denver averaged 12 drives per game during the regular season and only eight per game so far in the playoffs. I have no idea if the Broncos plan on snailing it with Seattle in the Super Bowl. If Denver does, though, that would be one legitimate reason for the Over/Under to be artificially low. But explanation #3 is a lot more intriguing.

  • George

    Just quickly chipping in on this;

    1) Weather – definitely interesting to know this, about the effects tending to be minimal. I typically stay away from totals where less than desirable weather is involved. Re: this and the recency bias – how many of the last six (I’m removing the Atlanta game) Seattle games were played in say 45 degrees or less. 4 out of the 6 according to the box scores and all 6 were below 50. I know this is taking team talent out of the equation (e.g. the slumping Seattle O, against the Arizona D in that game) but I think there is something there (I don’t have the statistics or trends to back it but just feel that there is something there). The talk about the over/under rising has also been attributed to the usual run of public money (people like seeing points being scored, people tend to take the over forcing the number up) and the fact that the weather isn’t going to be as bad as people thought has been considered as the reason people are having more faith in points being scored.
    2) Recency – I think this is 100% spot on. Unweighted figures for both teams has the game at give or take 56 points, weighted (assuming 95% as a consistent figure) has this at approximately 40 points. I guess they have tried to factor the weather in, as well as taking into account recent trends. As I think Seattle win (which leads onto point 3 which I hadn’t seen) I’d also consider the under more likely.
    3) I hadn’t thought of this but agree with the point. I just still can’t see how/why this line has shifted the way it has, I know there was talk of an anticipated buy back the other way (as couple of Vegas properties had stated that they thought that the game would go off as a pick ’em) but I really still can’t figure the angle (and I accept I could be very wrong and this could be one-way Denver). If anything it looks like it may go the other way (it shouldn’t cross 2.5).