2018 DVOA Projections

2018 DVOA Projections
2018 DVOA Projections
Photo: USA Today Sports Images

by Aaron Schatz

Here are our DVOA projections for 2018, updated from the season forecasts in Football Outsiders Almanac 2018.

For those new to our website, you can find an explanation of DVOA here. Note that there's a big difference between DVOA and projected DVOA. The DVOA ratings that appear on the website during the season are based on the actual play-by-play that happens during the season, with no future projection whatsoever. The numbers here are a forecast, with offense, defense, and special teams DVOA all projected separately using a system based on looking at trends for teams over the past decade. Our system starts by considering the team's DVOA over the past three seasons and, on offense, a separate projection for the starting quarterback. Then we look at a number of other variables which suggest when a team will be better or worse than would otherwise be expected due to standard regression towards the mean. Factors include major offseason personnel changes, coaching experience, recent draft history, combined tenure on the offensive line, and certain players returning from injury (or, in the case of these preseason updates, certain players getting injured in the preseason).

The numbers we are presenting here are exactly what the projection system spit out. As we say every year: "A few of them will look strange to you. A few of them look strange to us." As always, the offensive projections come out in a wider range than defensive projections because offense performance tends to be easier to predict (and more consistent from year to year) than defensive performance. If you are looking for subjective projections, Thursday we will be running our usual staff predictions article where we all talk about where we think the numbers are wrong.

We've also done our first playoff odds report simulation based on these updated DVOA projections, and I've added the playoff odds and Super Bowl championship odds to the table below. At the start of a new season, our simulation is very conservative about the average number of wins and losses expected for each team. Obviously, the NFL is going to have teams that are 11-5 or better, and it is going to have teams that are 5-11 or worse. It may seem like our simulation predicts half the league to go 8-8, but we know that won't happen. We also use a "dynamic" playoff odds simulation. Each time it plays through the season, it adds 1.5% to the DVOA of every winner and subtracts 1.5% from the DVOA of every loser before moving on to the next week's games. This reflects the fact that DVOA projections are just estimates, and actual performance during the season gives us better knowledge of how good or bad teams really are.

Personnel changes aren't the only difference between this updated simulation and the one we did for Football Outsiders Almanac 2018. This is a smaller simulation which only uses one set of mean projected DVOA ratings, rather than using 1,000 different sets of ratings to represent that some teams have a wider range of probable performance quality than others. The smaller simulation is even more conservative than the one from the book, so average win projections will come out a little closer to 8-8.

There are a number of suspensions, injuries, and holdouts impacting the start of the 2018 season. These are all accounted for in our weekly picks against the spread (part of FO Premium). They are not all accounted for in the DVOA ratings and simulation below. In the end, we only accounted for two backup quarterback situations. The penalties are lower than what we've used in previous seasons because it is early in the season and therefore the penalties, like the projected DVOA ratings, are closer to zero due to uncertainty.

  • In all simulations, Tampa Bay DVOA is -7.5% lower in Weeks 1-3 due to the Jameis Winston suspension.
  • In all simulations, Philadelphia DVOA is -5.0% lower in Week 1 due to the Carson Wentz injury. In the playoff odds simulation, we gave Wentz a 40 percent chance of returning Week 2, 60 percent chance of returning Week 3, 80 percent chance of returning Week 4, and 100 percent chance of returning Week 5.

The odds of getting the No. 1 pick listed below (and listed on the playoff odds report page) do not incorporate traded picks. Projected division champions are colored in light yellow and projected wild-card teams are colored in light purple.

PIT 18.8% 1 9.9 15.9% 1 -3.5% 8 -0.6% 17 -0.2% 21 0.2% 69.8% 12.0%
NE 18.0% 2 10.2 13.8% 2 -0.3% 15 4.0% 2 -4.1% 32 0.2% 77.4% 14.1%
NO 15.7% 3 9.4 12.7% 3 -3.2% 9 -0.2% 11 2.0% 4 0.4% 60.7% 8.4%
LAR 13.4% 4 9.3 2.8% 11 -7.2% 1 3.5% 3 0.3% 14 0.5% 58.5% 7.1%
MIN 11.5% 5 9.1 4.8% 7 -6.8% 2 -0.2% 12 0.2% 16 0.6% 53.3% 5.8%
PHI 11.1% 6 9.0 4.2% 8 -4.6% 3 2.3% 5 0.6% 10 0.8% 52.8% 5.7%
GB 10.3% 7 9.1 11.4% 4 0.5% 18 -0.5% 15 -1.0% 25 0.7% 53.2% 5.4%
DAL 7.5% 8 8.7 7.2% 6 0.2% 16 0.4% 8 0.0% 17 1.1% 47.8% 4.3%
ATL 4.3% 9 8.2 7.9% 5 3.0% 25 -0.6% 16 2.6% 2 1.6% 38.8% 2.9%
BAL 4.0% 10 8.4 -5.3% 24 -3.9% 6 5.4% 1 -0.1% 19 1.5% 42.5% 3.0%
SEA 3.0% 11 8.3 1.8% 12 -1.4% 13 -0.2% 10 0.3% 13 1.7% 39.7% 2.7%
HOU 0.8% 12 8.4 -2.4% 20 -4.5% 4 -1.3% 27 -2.9% 31 1.8% 44.0% 2.8%
WAS 0.2% 13 7.9 3.7% 10 1.7% 22 -1.8% 32 1.2% 8 2.4% 33.4% 1.8%
TEN -0.3% 14 8.2 0.8% 15 0.6% 19 -0.5% 14 -1.9% 26 2.0% 40.6% 2.5%
LAC -1.4% 15 8.0 0.9% 14 1.3% 21 -1.0% 22 -1.9% 27 2.6% 40.4% 2.2%
IND -1.7% 16 8.0 0.8% 16 4.1% 28 1.6% 6 -2.3% 29 2.3% 38.3% 2.0%
CAR -1.9% 17 7.5 -6.3% 27 -4.1% 5 0.2% 9 3.2% 1 3.1% 27.4% 1.3%
DET -2.0% 18 7.8 3.7% 9 6.6% 30 0.9% 7 0.6% 11 2.7% 30.0% 1.5%
KC -2.4% 19 7.7 -1.4% 19 3.8% 27 2.8% 4 0.5% 12 2.8% 34.8% 1.9%
CIN -2.8% 20 7.6 -0.9% 18 0.3% 17 -1.5% 29 1.6% 6 3.0% 29.2% 1.4%
CLE -3.2% 21 7.5 -5.1% 23 -3.6% 7 -1.7% 31 1.6% 7 3.2% 28.6% 1.4%
OAK -4.6% 22 7.6 1.1% 13 4.8% 29 -0.9% 19 -0.6% 24 3.5% 33.9% 1.6%
SF -4.9% 23 7.5 -0.9% 17 3.0% 24 -1.0% 24 -0.3% 23 3.8% 27.1% 1.2%
ARI -5.1% 24 7.4 -6.2% 26 -2.2% 11 -1.0% 23 1.2% 9 3.6% 24.7% 1.0%
JAX -5.5% 25 7.5 -6.4% 28 -1.8% 12 -0.9% 20 -0.4% 20 3.4% 30.1% 1.3%
MIA -7.2% 26 7.5 -4.7% 22 2.2% 23 -0.3% 13 -2.4% 30 3.7% 29.5% 1.1%
TB -7.2% 27 7.0 -3.0% 21 3.5% 26 -0.8% 18 2.6% 3 5.3% 20.6% 0.8%
DEN -7.8% 28 7.3 -9.5% 30 -2.7% 10 -1.0% 21 -0.3% 22 4.7% 28.6% 1.0%
CHI -11.2% 29 6.9 -9.2% 29 0.9% 20 -1.1% 26 0.1% 18 6.5% 18.3% 0.5%
NYJ -11.4% 30 7.1 -10.5% 31 -0.4% 14 -1.4% 28 -2.1% 28 5.8% 23.7% 0.8%
NYG -14.6% 31 6.4 -6.1% 25 6.9% 32 -1.6% 30 1.7% 5 9.3% 13.7% 0.3%
BUF -23.3% 32 5.6 -15.6% 32 6.7% 31 -1.0% 25 0.2% 15 15.2% 8.7% 0.1%

Which teams saw the biggest drop in mean wins since the book, and why?

  • Oakland, primarily due to the Khalil Mack trade but also other personnel changes.
  • Dallas, due to the illness that will keep Travis Frederick out of the starting lineup as well as other offensive line injury questions and cutting Dan Bailey.
  • Jacksonville, due to the Marqise Lee injury.
  • Arizona, based on a number of personnel adjustments that all changed small variables slightly.

Chicago has by far the biggest jump in mean wins since the book. It may not seem like it since the Bears are still near the bottom of the table, but Chicago had 6.2 mean wins in the book and is now at 6.9 mean wins. They went up 4.5% in projected DVOA, while no other team went up more than 1.2%.


50 comments, Last at 11 Sep 2018, 4:36pm

1 Re: 2018 DVOA Projections

Chicago is clearly ranked too low and Washington is ranked too high. My gut feelings are way better than this and I've had guts since before DVOA even existed! OH.SNAPPPPPPPPPPPPP

2 Re: 2018 DVOA Projections

You obviously meant to say that Denver is clearly ranked too low and Dallas is ranked too high. Evidently, someone swapped out your coffee for decaf this morning. *grin*

3 Re: 2018 DVOA Projections

I'm too lazy to check; how unusual is it to only have one team reach 10 mean wins, and zero teams reach 10.5 mean wins? It just feels weird to have 16 teams between 8 mean wins and 10.2 mean wins, and 5 teams at 7.5 mean wins, but then again it comports with my sense that the differences among all but the really bad teams are quite small, and random events will determine the final outcome.

35 Re: 2018 DVOA Projections


Miami might be okay.

Then there are the Jest and Buffalo. Buffalo is one Hue Jackson hiring away from going -2 and 18. That's a franchise that's trying to lose.

4 Re: 2018 DVOA Projections

Really interested to see the defensive alignment percentage for the Rams this year. With those two sluggers in the middle, it might be kind of fun to see the Rams play dime packages on 1st down a lot, even in the 1st half, and just dare teams to try to run the ball a lot.

(edit) To follow on, looking at their schedule, and with the usual injury caveats, I'll be mildly surprised if the Rams don't win 12 games, and shocked if they don't win 10. 9.3 mean wins just seems way too low.

7 Re: 2018 DVOA Projections

Aren't mean wins always conservative, in both directions? Look at the bottom of the chart. Someone will win fewer than 5 games this year, most likely--but mean wins doesn't know exactly who, so no one has a projection that low. Same thing happens at the top. Sure, the Rams look like they're sitting pretty in terms of schedule and likely wins--as do the Patriots, who haven't won under 10 since I was in high school--but mean wins is always conservative in terms of who will do the big winning.

Also, aren't mean wins basically a probabilistic measure? Say LA are 2:1 favorites for every one of their games (rough, I know): those 1/3 losses add up, even though eyeballing it they look amazing.

Edit: but I totally agree with your first point. They should try that at least occasionally, because that line is going to be bonkers.

11 Re: 2018 DVOA Projections

Sure, it is probabilistic, but I just don't think if you played the schedule 50k times the Rams would win fewer than 10 games half the time. I'm probably counting on (and I may well be wrong) Donald and Suh adding up to be greater than the individual players taken in isolation, due to so many o lines lacking sufficient interior talent to scheme effectively against two very dominant tackles.

15 Re: 2018 DVOA Projections

Winning 10 games half the time is hard.

For a team to win less than ten games less than half of the time, they'd have to have about a 60% chance of winning each individual game; so roughly 3/3.5-point favorites week in, week out. That's a tall ask, especially in a conference as competitive as the NFC.

There are four games this season where the Rams aren't favored by at least three points, per Vegas -- week 5 @ Seattle, week 7 @ San Francisco, week 9 @ New Orleans and week 13 @ Detroit, meaning they'd have to make that up in the other 12 games on the schedule.

Even if you accepted Vegas' odds as gospel -- and they're a bit overly favorable to the Rams, because that's where the money's going at the moment -- you'd still get an average of about 9.8 wins in the average season, based on historical winning percentages at various closing lines. 43 percent of the time, you'd expect them to finish 9-7 or worse -- and that's without taking into account the effect an injury to, say, Jared Goff or Todd Gurley would have on the lines going forward.

Winning 10 games half the time is ~hard~.

24 Re: 2018 DVOA Projections

People tend to underestimate the value of non-stars, especially in the context of a superior coach like Wade Phillips, who can use his dominant players in a way that allows his other players to be productive. The Rams, to me, are the most interesting team to watch this year, if you discount local rooting interests.

39 Re: 2018 DVOA Projections

I guess Ebukam and Longacre are their pass rushers in 2018? Last year they combined for 7.5 sacks and they weren't the primary pass rushers. It looks like Quinn and Barwin were the primary pass rushers. They combined for 13.5 sacks. It seems like with Suh upgrading pressure in the middle, that Ebukam and Longacre should be fine.

8 Re: 2018 DVOA Projections

My gut feeling is that this site, like other stats nerd sites in general, gets worse every year. Maybe this will be the year this place jumps the shark 2016 nate silver-style.

9 Re: 2018 DVOA Projections

Shocking news from the Steelers that Bell is planning to miss multiple games to start the season, apparently to protect his long-term health. I'm guessing the DVOA projection for them would change by a lot, right?

10 Re: 2018 DVOA Projections

KC's offense falls all the way from fourth to 19th, wow. But Jacksonville all the way down at 25th overall is the big surprise to me.

Amusing that the three New York state teams are projected to be the bottom three in the entire league.

12 Re: 2018 DVOA Projections

How much does Dickson factor into the projected ST rating for Seattle this year? I think that kid elevates them to a top 5 ST's unit.

13 Re: 2018 DVOA Projections

Brady's eventually gonna hit the age cliff; I think the odds are higher that it's this year than you guys do, apparently. Also think the proclamations that the whole AFC apart from Pittsburgh and New England stinks is typical of a backward-looking system, but there's a good chance that somebody (or even several teams) makes a big jump to the elite this year (arguably Jax did last year, with the point differential of a 12-win team)

18 Re: 2018 DVOA Projections

So, how did the projections do last year? I compared the 2017 projections to the actual results (by DVOA).

The standard deviation was 9.13 points. The three biggest whiffs:

New Orleans, predicted to be the #26 team in the league with a DVOA of -8.9%, actual result #1 team with 30.7% DVOA, difference = 39.6%

NY Giants, predicted to be the 7th best team with a DVOA of 6.5%, actual result #30 team with -22.3% DVOA, a difference of 28.8%

Minnesota, predicted to be the 20th best team with a DVOA of -2.9%, actual result the #4 team with a 25.0% DVOA, a difference of 27.9%

The three closest predictions:

Washington, predicted to be the 17th ranked team with a DVOA of -0.6% DVOA, actual result the 16th-ranked team with a DVOA of 0.6%, a difference of 1.2%

NY Jets, predicted to be the 32-ranked team with a DVOA of -18.6%, actually the 26th-best team with a DVOA of -17.2%, a difference of 1.4%

A tie between Atlanta and Buffalo, both off by 2.1%

In total, 9 of 32 team predictions were within 5% of DVOA of the actual result. 6 were off by 15% or more (and Denver off by 14.5%)

23 Re: 2018 DVOA Projections

It really is just a wild guess with regard to rookie production, so when teams like the Saints or Vikings get really good play from rookies, in the units that were the biggest flaws the year previously (Saints with dbs, Vikings with o-linemen), the projection is going be substantially off on the low side. Accurate predictions are hard to make. Especially about the future.

26 Re: 2018 DVOA Projections

The predictions are very compressed. The NFL has a lot of year-to- year varuability so It is hard to figure out which team will be best and which will be worst. If you think Pittsburgh is going to be the best team this year then it is very very likely they will end up with a DVOA of more than 19% and more than 10 wins. Also if Buffalo turns out to be the worst team they will have a fewer wins and a worse DVOA. Over the last 5 years the average end of season DVOA of the best team was 33% and for the worst team it was -31%. I think the most useful way to look at these predictions is that FO "thinks" that Pitt or NE will be the best team (13 wins & DVOA at 30%) but due to uncertainty they will most likely be above average (10 wins & 18% DVOA).

27 Re: 2018 DVOA Projections

I think it would be useful to communicate this better to include the error bar in the calculation; my guess is that it's about +/- 10% on DVOA & +/- 2 wins. The size of those error bars really lend some perspective to the comment "but someone is going to win 12+ games in the AFC".

21 Re: 2018 DVOA Projections

Jacksonville, Kansas City offenses? Panthers dropped from 17nth in offense last year to 27nth. They must all have had horrendous off-seasons.

32 Re: 2018 DVOA Projections

Well, KC has Cam Erving starting on their line right now, so that's got to be part of it...Mahomes has to also be a bit of a question mark still, too, especially compared to the enigmatic-but-established current Skins QB.

Jax has to be going through a regression to the mean; being able to hide Bortles' weaknesses last year isn't really sustainable this year. The odd thing is that adding Fournette, changing scheme accordingly, etc. didn't really increase their rush YPG much - might be down-and-distance related?

48 Re: 2018 DVOA Projections

I don't know in what universe this season's Jaguars defense isn't projected to be in the top 10, but I seem to have wandered there by mistake. My personal opinion is that the team is ten times as likely to finish in the top three in the entire league rather than outside the top three of the AFC South, which is where they're apparently projected, especially given Houston's offensive line woes and Indy's efense (sic). But #BlameBortles or whatever.

28 Re: 2018 DVOA Projections

What is the cause for the mild "optimism" with regard to NE's defense? What factors does DVOA see that makes it think NE will jump up to the middle of the pack?

31 Re: 2018 DVOA Projections

I'm going to guess their 4 games against the Bills and Jets ranked 31 and 32 in offense. Historically football outsiders has demonstrated defense variance is higher year to year than offense. It's not that unusual to move from dreadful on defense to middle of the pack in one year.

33 Re: 2018 DVOA Projections

As I understand it, games against subpar competition can elevate traditional stats, but DVOA will still recognize the flaws. You make a good point about the variance, but wouldn't that just mean that bigger changes are allowed? Shouldn't there still be some factors that lead DVOA to believe a particular team is a good candidate for a big change?

34 Re: 2018 DVOA Projections

That's a correct assessment of DVOA (as opposed to VOA). Some of this has to be regression to the mean; NE had about 5 games last year that were historically bad on defense from a DVOA POV - it seems unlikely that those outliers will be repeated. I'm not sure I buy the regression argument that underpins this in the case of the Pats as the front "7" seems like it might be really lousy again this year, but we'll see.

37 Re: 2018 DVOA Projections

Data nerd complaint: when you change team monickers, it can really mess with stuff.
I like the new ones better, but consistency would be nice. Thanks

41 Re: 2018 DVOA Projections

I hate "JAX" as an abbreviation. It is the only abbreviation that uses a letter that doesn't actual exist in the city or team name. But I realize it is a common abbreviation.

I am happy about the change to LAC and LAR, because there was no reason to have the 4-character abbreviations.

I think the important thing about these abbreviations is consistency, so that you can usually know the abbreviation without looking at it. Many outlets do not have consistency. For some reason, some places will use TAM as the abbreviation for Tampa Bay, but use SF for San Francisco. Since all the 2-word cities just use the first letter of each word, why would Tampa Bay be TAM? Then some places will use GNB or GBP for Green Bay.

I know some sites like to use 3-characters for all teams. That's fine, but be consistent. If New England is NWE, then shouldn't New Orleans be NWO instead of NOR?

47 Re: 2018 DVOA Projections

NFL.com has been using just "LA" for the Rams but "LAC" for the Chargers on their upper score bar for each of the past two years, which kind of comes off like the league is treating the Chargers as the stepchild team in that city.