Cowboys Lead Way Too Early 2022 DVOA Projections

Dak Prescott, Dallas Cowboys
Dak Prescott, Dallas Cowboys
Photo: USA Today Sports Images

NFL Super Bowl - Last year, I presented a new set of way, way too early DVOA projections right after the Super Bowl. It's time to do it again for 2022!

Once again I remind readers not to get too excited: the goal here was to be incredibly simple. That being said, last year's way too early projections were closer to reality than you might expect. Eleven of this year's playoff teams finished in the top half of the way too early projections. The exceptions were New England, Philadelphia, and Cincinnati.

First, I adjusted 2021 DVOA ratings based on games where teams sat starters or had COVID issues or used backup quarterbacks. In the table below, I listed the quarterback who was considered the starter for each team. For example, the New Orleans offensive DVOA below is based only on games with Taysom Hill at quarterback. Seattle offensive DVOA is based on a rating without Weeks 6-8. The Cincinnati ratings don't include Week 18. And so on.

Next, I used very simple equations that project next year's DVOA based solely on last year's DVOA and the usual regression to the mean. Defense and special teams regress to the mean stronger than offense does.

Then I made a few changes based on roster change and development that we know is coming:

  • I boosted Cincinnati a little bit. Normally my projections are all based on regular-season stats, but with Cincinnati we now have a four-game sample in the playoffs that shows that they played better than they did during the regular season.
  • I gave a small bonus to teams with second-year starting quarterbacks except for New England (already pretty high up in DVOA) and San Francisco (since Lance doesn't have much experience yet).
  • I slightly improved the Baltimore defense based on all the injuries they suffered in 2021.
  • I penalized New Orleans because they are by far the team with the most salary cap difficulty for next year, even if they restructure a number of contracts.
  • I manually set the Tampa Bay offense at replacement level, i.e. the average backup, before doing the equation that regressed 2021 performance to the mean.

Then I re-normalized all the DVOA ratings so the league averaged 0.0%.

That's it. I didn't account for all the things that are accounted for in the usual preseason projections, such as three-year trends, coaching changes, turnover regression on defense (as opposed to general regression), and free-agent movement. I didn't account for any injuries except for starting quarterbacks and Baltimore's defense. I didn't account for salary-cap issues except for New Orleans. I didn't account for any possible quarterback trades, or where Deshaun Watson or Jimmy Garoppolo might be playing next year.

The goal here is basically just to see who looks best for 2022 based mainly on quarterbacks returning from injury and offense being more predictive than defense. The result looks a lot like the 2021 DVOA ratings with a few changes. That means that differences between DVOA and conventional wisdom will be differences here. Dallas, New England, and Seattle are higher than conventional wisdom. Cincinnati, Tennessee, and Atlanta are lower than conventional wisdom.

The big difference between this year and last year is that I also compiled schedule strength based on the way too early projections. The defending champion Rams have the hardest projected schedule. The NFC South is a weak division, but the entire NFC West is in the top 10 and the AFC West is a tough division. The Rams' other opponents based on finishing in first place are Dallas, Green Bay, and Buffalo: the top three teams in the way too early projections. The Arizona Cardinals also have a particularly tough projected schedule. Every other team has an average opponent projection of 2.0% or lower.

The easiest projected schedule belongs to the Baltimore Ravens. They play the AFC East and NFC South, plus Denver, Jacksonville, and the New York Giants. The Dallas Cowboys also stand out as the only team in the top 10 with a bottom 10 schedule.

TEAM QB TOT
DVOA
RK OFF
DVOA
RK DEF
DVOA
RK S.T.
DVOA
RK SCHED RK
DAL D.Prescott 15.1% 1 8.6% 5 -6.0% 2 0.6% 7 -1.7% 29
BUF J.Allen 12.3% 2 5.4% 9 -7.1% 1 -0.2% 19 0.5% 14
GB A.Rodgers 11.2% 3 13.0% 1 0.7% 19 -1.1% 32 -0.5% 22
NE M.Jones 10.7% 4 5.9% 8 -5.0% 3 -0.1% 18 0.0% 18
SF T.Lance 10.3% 5 8.2% 6 -2.7% 7 -0.7% 26 0.5% 13
KC P.Mahomes 9.9% 6 10.6% 2 1.9% 25 1.2% 3 1.1% 8
LAR M.Stafford 9.9% 7 5.9% 7 -3.2% 5 0.8% 5 3.5% 1
SEA R.Wilson 9.8% 8 10.5% 3 1.4% 23 0.7% 6 -0.4% 21
LAC J.Herbert 6.4% 9 9.2% 4 2.0% 27 -0.7% 28 0.1% 17
ARI K.Murray 5.2% 10 2.3% 13 -3.2% 6 -0.3% 20 2.7% 2
IND C.Wentz 4.2% 11 2.4% 12 -1.9% 8 0.0% 14 -1.7% 28
BAL L.Jackson 3.1% 12 2.1% 15 0.8% 20 1.9% 1 -3.6% 32
CIN J.Burrow 2.9% 13 2.6% 11 0.5% 18 0.8% 4 -0.8% 24
CLE B.Mayfield 2.5% 14 2.2% 14 -0.9% 10 -0.6% 25 -2.6% 31
PHI J.Hurts 2.5% 15 4.5% 10 2.0% 26 0.0% 15 -1.3% 26
MIN K.Cousins 1.8% 16 2.0% 16 0.2% 15 0.0% 13 0.8% 9
TEN R.Tannehill -1.8% 17 -2.2% 18 -0.8% 11 -0.4% 22 0.5% 12
LV D.Carr -2.6% 18 -1.8% 17 0.4% 16 -0.4% 21 1.7% 5
CHI J.Fields -2.7% 19 -3.6% 22 -0.3% 12 0.6% 8 0.3% 15
MIA T.Tagovailoa -4.1% 20 -2.9% 20 0.4% 17 -0.8% 29 1.8% 3
WAS T.Heinicke -4.6% 21 -2.4% 19 2.4% 28 0.2% 12 0.2% 16
NO T.Hill -4.6% 22 -8.4% 29 -3.9% 4 0.0% 16 -1.6% 27
TB B.Gabbert -6.0% 23 -6.5% 26 -1.3% 9 -0.7% 27 0.6% 10
NYG D.Jones -6.8% 24 -5.9% 24 1.2% 21 0.3% 11 0.6% 11
DEN D.Lock -8.8% 25 -6.5% 25 1.3% 22 -1.0% 30 1.7% 4
NYJ Z.Wilson -8.8% 26 -4.4% 23 5.8% 32 1.5% 2 1.4% 7
JAX T.Lawrence -9.1% 27 -3.3% 21 4.8% 31 -1.0% 31 -1.3% 25
HOU D.Mills -9.7% 28 -8.5% 30 1.5% 24 0.3% 9 -0.4% 20
CAR S.Darnold -11.0% 29 -10.5% 31 0.0% 14 -0.5% 24 -2.6% 30
PIT M.Rudolph -11.8% 30 -11.8% 32 -0.2% 13 -0.1% 17 -0.2% 19
DET J.Goff -12.4% 31 -8.3% 28 4.4% 29 0.3% 10 1.6% 6
ATL M.Ryan -13.2% 32 -8.2% 27 4.6% 30 -0.4% 23 -0.7% 23

Comments

34 comments, Last at 23 Feb 2022, 12:47pm

4 Seattle's offensive DVOA

Here's Seattle's offensive DVOA by week this year. The splits are rather remarkable. Listed is the rank of where that rating would have ranked for the entire season.

Wk 1-5 31.5% 1
Wk 6-8 (Smith) -13.0% 25
Wk 10-13 -19.3% 29
Wk 14-18 32.8% 1

6 I wonder if these…

I wonder if these projections would get more accurate if you put greater weight on the second half of the season, or some other later-season grouping of games. I know Bill James has found that baseball teams with good second-half records are more likely to improve in the following season.

8 Also, DVOA likes Seattle's…

Also, DVOA likes Seattle's offense relative to public perception, because they ran the fewest plays in the league (so they were actually efficient from a play-by-play metric) and they avoided turnovers. Like in Week 3 vs Minnesota, they only scored 17 but DVOA loved their performance as they combined fast scoring drives in the 1st half with 3-and-outs in the 2nd half.

3 The Cowboys botched the…

The Cowboys botched the number one seed in 2007 and 2016 by not winning a playoff game, and last season they botched the number one DVOA rating by not winning a playoff game. Can they do it again next year? Tune in for another season of Jerry's World to find out!*

 

*Viewer Discretion Advised: Jerry's World takes no responsibility for disappointment due to sloppy execution, underperforming RBs, or penalty prone play

9 Deshaun Watson

What is TB's projection if Deshaun Watson is traded to them and able to start the whole season?

12 I tend to think the Ravens…

I tend to think the Ravens are the prime candidate for a bounce back after a year in Injury Hell. Still a great coach and “maybe” the best qb in division. It’s still at least debatable. 

26 I would take Burrow also,…

I would take Burrow also, but not trust the Bengals organization as a whole.  Will they properly scout and draft/buy free agents to fix the offensive line?

The 17 game schedule has made it so that division rivals now have 3 non common opponents.

While BAL plays NYG, JAX and DEN, CIN plays DAL, HOU, KC.   BAL may get 2 game lead just from that advantage.

30 "Lamar Jackson has been…

In reply to by RickD

"Lamar Jackson has been sliding. (-2.9 DVOA in 2021)  I'd definitely take Burrow over him at this point"

Burrow's best season is a paltry 5.1% DVOA, while Jackson's 34.9%. If you build a strong O-line around Jackson, his upside is unanimous MVP. Not sure what Burrow's upside is, but I doubt MVP. 

13 Kind of interesting to see…

Kind of interesting to see the NFC strongly represented in the top-5 and top-10 since I've found myself thinking that the NFC is wide open compared to the AFC, where a stable upper crust seems to be emerging behind established young QBs. Then again, this projection assumes Rodgers and Wilson stay in place, and my perception of a stronger upper tier in the AFC might prove more true over the next 2 to 5 years than just next season.

15 I penalized New Orleans…

  • I penalized New Orleans because they are by far the team with the most salary cap difficulty for next year, even if they restructure a number of contracts.

It's not that dramatically different from the Cowboys. The difficulty is that you're only looking at it on a 1-year basis. The fact that the Saints are like, $80M over for next year's immaterial, they obviously are going to restructure some players and push money forward (I mean, they have to).

On a 2-year basis the Saints have -$45M, and the Cowboys have -$17M, but literally two cuts (Roby and Jenkins) would bring the Cowboys/Saints to essentially equal on a two-year basis.

Both the Saints and the Cowboys literally have more people under contract for next year than can actually make the team. The huge negative cap values are basically just the team's way of saying "we don't know who we're going to keep yet."

That's not to say the Saints don't have cap difficulty. Of course they do - their difficulty is that they're in a similar position to the Cowboys except their team sucks.

16 Hindsight is 20/20, but the…

Hindsight is 20/20, but the Cowboys looked to be very inflated DVOA wise last year because they beat up on terrible teams. In the exact opposite way that the Rams under McVay sat their starters for the entire fourth quarter twice, once versus the Texans, again versus the Jags, the Cowboys just kept driving up the score. As a result I never thought the Cowboys were really the best team in the league last year, and I don’t think I’m alone in that.

Combine that with the salary cap issues, and I’m predicting a slightly above average team next year.

25 Cowboys

The Cowboys probably won't be quite as good next year due to defensive regression and a cap casualty or two, but if you go by the typical Cowboys result -- losing to good playoff teams -- the only teams that they are very likely to lose to on their schedule are the Packers, Rams, Titans, and Bengals. Throw in a random division loss and they're 12-5 again, also likely with the third seed again (Packers and an NFC West team will have the first two seeds, and I have no idea who wins the South, but they won't get to 12 wins unless Watson or Wilson go to Tampa).

Next season doesn't look like it'll be much different than this one for the Cowboys in terms of results.

27 That still leaves them the…

That still leaves them the best team in the NFC East. PHI snuck into the playoffs this year on a cupcake schedule that is unlikely to repeat. WAS doesn't have a quarterback and NYG is still a mess.

The defensive regression is likely going to hit them hard though. Their defense drastically improved from 2020 to 2021, and did so by forcing the most turnovers in the league. That's unlikely to repeat next year; in 2014, the Cowboys forced the most turnovers in the league, the next year, they forced the fewest turnovers in the league!

17 Is there data to back this up?

I gave a small bonus to teams with second-year starting quarterbacks except for New England (already pretty high up in DVOA) and San Francisco (since Lance doesn't have much experience yet).

I am wondering.  Are this adjustment and the exceptions to it backed up by data? Do we have reason to believe — from the data — that bad second year starters get better, but good ones don’t? Or that second year QBs with only a little playing time do not improve?

edit: before anyone jumps down my throat, by “Bad” I do not mean “destined to be bad forever”. Just bad by DVOA/DYAR in their first year.

18 Do we have reason to believe…

Do we have reason to believe — from the data — that bad second year starters get better, but good ones don’t?

The typical jump for first-to-second year QBs is huge, like 13% DVOA. So obviously, yeah, "good" QBs don't improve nearly as much on average or else we'd have a lot more Aaron Rodgers types.

19 Isn’t that just regression…

Isn’t that just regression towards the mean, though?

It sounds like there were three “adjustments” going on here. First: offenses and QBs regress toward the mean. Second: second year QBs are assumed to improve by X. Third: Jones and Lance’s offense get an extra adjustment down because… reasons?

I’d believe there is data to support the first two adjustments. I fear the third one was “we didn’t think the numbers looked right, so we changed them”. 

That’s why I wondered whether there was data to support QBs with a good first year DVOA not being expected to improve much. Or QBs with few first year passes.

21 First: offenses and QBs…

First: offenses and QBs regress toward the mean. Second: second year QBs are assumed to improve by X.

The right way to do it would be to regress first-year QBs to their mean. As in, Mac Jones had a 6.8% DVOA. That's way high for a first-year QB and likely an outlier. If he would be a first-year QB again next year, he'd probably do much worse. So your Bayesian estimate for Jones pulls him down. Lawrence was pretty typical for a first-year QB, so don't bother. Darnold, Fields, and Wilson were pretty bad, so you boost them up a bit.

That's what regression to the mean is - you don't expect them to change, you just expect that your observation was biased upwards by chance because your 'base expectation' for a first-year QB is "bad", with some spread. Jones's first year pulls that expectation up, but the prior pulls it down.

Now you boost all second year QBs by the expected improvement. But note what happens: the net change is that Jones got a downward adjustment (due to regression) then an upward adjustment (due to expected improvement). So basically no shift.

Functionally it basically works out the same. Jones was a clear first-year outlier, just like Herbert was. And yes, OK, Herbert did improve quite a bit from year 1 to year 2, but it's pretty clear he's on a path to "elite QB" at this point, which by definition is a far outlier, so it's OK to be cautious about believing in those guys.

23 I hear you. You think that…

I hear you. You think that first year QBs regress towards some fairly large negative DVOA before you can add in a second year improvement. It’s a theory, and I can see why you would think it.

But is there any data to support it? Is it really true that, on average, first year QBs with a lot of passes (say, >300) and solidly positive DVOAs (say > 5%) don’t improve much in their second year? Less than you would expect by the normal regression towards 0% and the average second year improvement?

That’s what I want to know. Was this an “eyeball” adjustment, or based on data?

28 You think that first year…

You think that first year QBs regress towards some fairly large negative DVOA

No, that's the wrong way to think about it. You're talking about regression to the mean as something that's forwardly active. It's in the words you use: "regress towards," present tense. It's something that's happening to them.

That's not how regression to the mean works. Regression to the mean is a Bayesian-type process. You're given a coin. You start off assuming that a coin is fair, because those are the only coins you regularly see. You flip it 4 times, get 4 heads. The measurement says that the coin's "heads rate" is 100%, but you don't believe it, because you super-rarely see trick coins, and 4 flips is small statistics with huge error bars. But you're not confident of that, so your "new" expectation is a little higher than 50%, but definitely not 100%. Your "100%" expectation "regressed back" to its expected mean. But that's a terrible phrasing. The better phrasing is "you updated your 50% expectation to something higher, but not all the way to the data." Your base expectation for Mac Jones was like -20% DVOA, he's ludicrously higher, so you update your expectation for his ability level slightly above -20%. 

It's the same thing here. Asking "is there any data to support" the fact that you should regress a first-year QB to a large negative DVOA? Yes. The fact that first-year QBs have, on average, a large negative DVOA! That's their mean! That's what you regress your evaluation based on his first-year data.

Your beginning expectation is "first year QBs are bad." Then you get data. In Mac Jones's case, it was way above expectations. Because that data's noisy, your new updated expectation is still lower. Because first year QBs are bad. 

Is it really true that, on average, first year QBs with a lot of passes (say, >300) and solidly positive DVOAs (say > 5%) don’t improve much in their second year? Less than you would expect by the normal regression towards 0% and the average second year improvement?

Regressing first-year QBs to a negative mean is just math. That's their mean, first year performances are noisy, so you reduce performances above and boost performances below. The problem with trying to back it up with actual historical data is that the sample size is already small enough, reducing it further's just going to make things noisy. 

That’s what I want to know. Was this an “eyeball” adjustment, or based on data?

It's obviously an eyeball adjustment, that's why Aaron just named the two QBs it didn't affect. Literally all of the bits and pieces adjustment (for example, the Saints cap adjustment) are just eyeball adjustments. But it's totally a justified thing to do.

I just don't understand what you're expecting here. The second year adjustment for QBs is huge. It's literally the only year you can see a statistically significant jump in performance. You really think the base expectation for New England's offense, with a second year QB, should be a top-5 QB? I mean, you can't be saying that, right?

You've got to just be figuring that he's likely to improve some, but not quite as much as the others. And if you're doing a handwavy "way too early" thing here, "zero" is close enough to that anyway given how much of an outlier he was historically.

It should be noted I don't even think Aaron's giving the full "second year boost" - last year he called it "small." Which actually makes sense, because the typical second year QB doesn't actually improve as much as the average second year boost, since many second year QBs might not get a long enough career to fall into that "typical boost."

Again, it's just about expectations. You can't set your expectations for Jones to be much higher than last year. There's just not enough data to expect him to be that high. I mean, the other thing you could do is take Jones's first year, give him the second year boost (so now he's crazy high) and regress to the expectations of a second year QB. And Jones would get a much higher downward adjustment than the other guys. Again, same overall effect. "No change" is perfectly reasonable.

22 Timing was weird

The timing was weird to do an XP post because it was sort of announced, but not announced, a while ago. Then finally made official this week.

24 I'm all about the Bills…

I'm all about the Bills being projected to have the #1 Defense again, but I'm skeptical of the forecast. Currently we have about 3000 defensive snaps as free agents and we're over the cap, so definitely not all of those dudes are coming back.

33 I don't think the bills will…

I don't think the bills will be #1 in yards and points again, but I also don't think the bills were the best defense in the league and so I think the bills d can be as good or better next year without being #1 in yards or points or dvoa or whatever.  Even if the bills lose jerry Hughes, Harrison Phillips, and Levi Wallace, I think a combination of improvement from young guys like the DEs drafted in '21 and ed Oliver, getting Tre white back, drafting a DB or DT or LB, could all make up from any FA losses/veteran declines and result in a defense as good or better next year. So consider me optimistic on buffalos d next season!