Sixteen years of NFL over/under trends hold predictive powers

Ethan Miller/Getty Images

The 2018 NFL season is set to begin Thursday night. That leaves just a couple of days for sportsbooks to make final adjustments to their season win totals (over/unders) and fans to make final decisions about which teams they believe will go over or under those opening lines.

The New England Patriots' over/under is 11 wins, putting them atop the win-total board before start of the season for the third consecutive year and ninth time since 2004; they've been alone at the top on five of those occasions. The Buffalo Bills project to win 5.5 games -- the fewest in the NFL -- producing one of tightest ranges this century between the teams with the highest and lowest over/unders.

But does any of this really matter? I looked back at NFL teams' over/unders, regular-season wins and over/under results since 2002, which was the start of the 32-team era. I paid special attention to three topics: (1) specific over/under levels; (2) specific franchises; and (3) teams' projected improvement or decline from one season to the next.

Some of the findings were as expected. The Patriots, Indianapolis Colts and Philadelphia Eagles have frequently gone over their win totals, for example, while the Cleveland Browns, Chicago Bears, Jacksonville Jaguars, Los Angeles Rams and Oakland Raiders have often gone under theirs. Other trends required a little more digging.

Win totals for the 2018 season are from Westgate Las Vegas SuperBook, as of Sept. 3. Historical win totals are sourced from Sports Odds History.

Teams projected to be below average trend over. Teams projected to be average or a little above trend under.

First, it's important to keep in mind that over/unders, like preseason Super Bowl odds, are highly correlated (0.82) with teams' wins in the previous season -- and only modestly correlated (0.45) with how teams ultimately perform in the season that's being projected. In that way, over/unders are essentially adjusted versions of the prior season's standings, often reflecting some upward or downward movement toward eight wins, the NFL average.

Teams loosely perform as their over/unders suggest they might, but we otherwise can't pin down what a team's real number of victories will be based on its over/under (and why watch or wager if you could?)

There are nevertheless some subtle trends lurking in the midst of all this data -- particularly when we look at specific over/under levels. Since 2002, NFL teams handicapped to earn six to 7.5 wins have gone over 85 times (51.5 percent), under 72 times (43.6 percent) and pushed eight times (4.8 percent). Most of that trend toward the over came by way of teams projected to win six or 6.5 games, which this season includes the Browns (six), Tampa Bay Buccaneers (6.5), Arizona Cardinals (six), Miami Dolphins (6.5) and New York Jets (six).

Meanwhile, teams projected to win eight to 10.5 wins have done the opposite -- they've gone over 124 times (44.3 percent), under 142 times (50.7 percent) and pushed 14 times (5 percent). In short, teams expected to be a little below average have trended slightly over, while teams expected to be average or a little above that mark have trended slightly under. On balance, reversion toward the NFL's eight-win average has won out.

Sample sizes get smaller and results get muddier at the extremes, but Patriots fans might like to know that 11 of the 18 teams (61.1 percent) handicapped at 11 or 11.5 wins since 2002 -- as New England is this season -- have gone over, including the Pats themselves on three of five occasions. Those same Pats fans should hope that doesn't mean they're due to go the other way in 2018.

Some teams are more consistent than others.

Recent franchise history also bodes well for the Patriots. In the past 16 NFL seasons, New England has gone over its win total 12 times and pushed once. Sportsbooks forecasted some year-over-year decline for the Pats in 14 of those seasons, but Tom Brady, Bill Belichick & Co. have rewarded over bettors time and time again. The shoe will eventually drop, I suppose, but if it's after Brady retires, it will likely be baked into the line.

The Colts and Eagles each hit the over 11 times since 2002, second-best behind the Pats. But the two teams have met very different fates of late. The Colts have gone under their total for three consecutive seasons, while the Eagles went over theirs in four of the past five seasons (they also won Super Bowl LII, of course). The Houston Texans, Denver Broncos and Pittsburgh Steelers have each gone over 10 times, but they've done so relative to very different over/under averages: 7.6 for Houston, 8.8 for Denver and 9.5 for Pittsburgh.

The Browns, as you might expect, have done almost the exact opposite of the Patriots. Cleveland's over/unders suggested that the team would improve in 13 of the past 16 seasons, if only to a modest average of about six wins. The Browns instead averaged fewer than five wins, going under their win total 12 times in 16 years.

Teams projected to improve trend over. Teams projected to decline trend under.

Despite winning zero games last season, Cleveland has seen its win total bet up to six games in 2018, and the team is attracting attention in Las Vegas and on HBO's "Hard Knocks." That six-win difference represents the highest projected year-over-year win increase before the start of a season since at least 2002 -- admittedly an easier feat when coming off a base of zero.

So how do NFL teams do when their over/unders suggest they'll win more games in the upcoming season than they won during the previous season -- that is, when they're expected to improve like the Browns are? And how do teams do when they're expected to decline, like this season's Vikings (13 wins in 2017, O/U 10 in 2018) and Bills (nine wins in 2017, O/U 5.5 in 2018)?

Since 2002, teams expected to improve by at least one victory from the previous season have gone over their win total 101 times (53.4 percent), under 79 times (41.8 percent) and pushed nine times (4.8 percent). Most, in other words, improved by more games than their over/unders implied they would. There is considerable overlap between this group and the "below average" win-total teams we looked at earlier, so it makes sense that both trend toward the over.

Meanwhile, teams projected to decline by at least one win have basically done the opposite: They've gone over 71 times (40.1 percent), under 95 times (53.7 percent) and pushed 11 times (6.2 percent). This group has a great deal of overlap with the "above average" win-total teams we reviewed earlier, and it follows that both trend toward the under.

But remember: While these subtle trends might make for good podcast conversation, they'll rarely be enough to overcome the heavy vigs and careful pricing that sportsbooks attach to win-total wagers. Most of this season's over/unders carry negative odds (-110 to -170), and those bets have an average vig of over 30 percent (-131). Keep that pricing in mind as you peruse the board and ponder whether the Browns will finally rise or the Patriots will finally fall.

Greg Guglielmo is a freelance writer based in New York. His work appears on FiveThirtyEight and his blog, ELDORADO. You can find him on Twitter @eldo_co.