Week 1 usually brings a lot of upsets to the NFL since nobody knows for sure just how good teams will be in 2018. Despite all the advanced statistics we have, even our Football Outsiders preseason projections are going to be off a good amount.
Unfortunately, the same things that make it tough for Vegas to set the odds also make it tough for us to figure out which odds are most likely to be incorrect. Many of the games this week have very small lines, with only three games having a line of more than a touchdown.
Based on what we know from the FO preseason projections, these are the most likely upsets with lines of at least three points:
Cleveland (+3.5) vs. Pittsburgh
Dallas (+3) at Carolina
New York Giants (+3) vs. Jacksonville
Oakland (+4) at Los Angeles Rams
Seattle (+3) at Denver
For our first Upset Watch of the year, we will revisit Football Outsiders' most surprising preseason projection, and look at why the Giants are a better bet to beat Jacksonville than most people realize.
Upset Watch: New York Giants (+3) vs. Jacksonville Jaguars
You know that Jacksonville had a stellar defense last year. You may not know that the Jaguars had an average offense, not a bad one. But what matters is looking at how they put together an average offense, because it's tied in some ways to that defense.
The Jacksonville offense was excellent at getting out to fast starts last season, ranking No. 1 in offensive DVOA in the first quarter, then 21st the rest of the game. Getting out to fast starts and then having a stellar defense to protect the lead allowed it to dictate the game script in most of its games. They could put the game in the hands of the ground attack and keep Blake Bortles out of difficult situations. As a result, the Jaguars went 10-2 when they scored first, but 2-5 when the opponent scored first (including the playoffs).
It is unlikely that the game scripts will be this favorable for Jacksonville this year.
First of all, there is no year-to-year correlation whatsoever when it comes to being particularly strong on offense in the first quarter, meaning the Jaguars aren't going to get out to as many leads. Second, they aren't likely to be as good at protecting those leads. As much talent as the Jaguars have on defense, there simply is more year-to-year variation in defensive performance than offensive, and more reason to believe that the Jaguars may field a good defense in 2018 rather than a great one.
One of the most interesting findings in our team projection system is that run defense tends to predict the next year's total defense better than pass defense does. That's a problem for a team such as Jacksonville, which was so much better at pass defense (No. 1 in DVOA) than it was in run defense (No. 26 in DVOA). Since 1988, 15 teams ranked in the top five for overall defensive DVOA but were also 15th or lower against the run. Only four of those teams had a top-five overall defense the following year, and on average they fell to 11th.
(The caveat here: Jacksonville ranked 31st against the run in the first half of the season but 13th in the second half after trading for Marcell Dareus. In general, first-half/second-half splits are surprisingly meaningless when it comes to predicting performance the following year, but perhaps this one is more meaningful because it is tied to the acquisition of a specific player.)
There are other reasons to believe that the Jacksonville defense will not play at quite the stellar level of 2017. There's the Plexiglass Principle, the idea that teams that improve significantly from year to year will regress a bit in the third year. There's the question of turnover ratio, which tends to considerably regress toward the mean from year-to-year; Jacksonville was third in takeaways per drive last season behind only Baltimore and Detroit. And there's the issue of health, as Jacksonville was the second-healthiest defense in the league based on our adjusted games lost metric (trailing only Minnesota).
The Jaguars are likely not going to be as healthy this year, but that doesn't matter in Week 1 when most injuries haven't happened yet. No, health is really in issue on the other side of the ball, where the Giants' offense was decimated by injuries last year. They finished 28th in adjusted games lost and most importantly lost stud wide receiver Odell Beckham Jr. for most of the year. Now the Giants have Beckham back, a healthy Sterling Shepard, rookie running back Saquon Barkley and some changes on their offensive line that have them expecting big things from the offense. We don't expect quite as much success as the Giants do, but we do expect improvement.
We may be lower on Jacksonville than anyone else in the football commentariat, but we still think the Jaguars are better than the Giants. Nonetheless, consider the likely improvement of the Giants' offense, the likely decline of the Jaguars on both sides of the ball and the fact that this game is at MetLife Stadium -- and you can see the storyline for a classic Week 1 upset.
Cover Watch: Dallas Cowboys (+3) at Carolina Panthers
The Cowboys were 9-7 last year, but reading commentary about them on Twitter, you are forgiven for thinking maybe they were 6-10 going on 1-15. This is not a terrible football team.
The Cowboys are just a year removed from going 13-3, and last year they ranked eighth in the Football Outsiders DVOA ratings at midseason. Then came the Ezekiel Elliott suspension and, more importantly, injury to left tackle Tyron Smith. Yes, the Cowboys will be missing center Travis Frederick for part of this season because of illness, but Smith and Elliott are back and the Cowboys' offense should be improved over its lousy performance from the second half of last year.
Meanwhile, Carolina is staring into the heart of regression when it comes to serendipity. The Panthers rode into the playoffs last year on the back of a fairly unrepeatable performance in close games: 8-1 in games that involved a fourth-quarter comeback or a game-winning opportunity on either side of the ball. They also got some extra takeaways thanks to luck recovering fumbles: nine of 14 on offense and 11 of 14 on defense. (Fumble recovery tends to be random and usually around 50-50.) Norv Turner may be an improvement coordinating the offensive side of the ball, but all else being equal, units tend to take a small step back when they have to learn new schemes with new coordinators.
Combine these issues, and Dallas ends up eighth in the Football Outsiders preseason projections, with Carolina down at 17th. ESPN FPI has them closer, with Carolina at 11 and Dallas at 13. But that still indicates a good chance of a Cowboys upset (or really close cover) on the road this Sunday.