Buffalo Bills Back to No. 1 in DVOA

Taron Johnson, Buffalo Bills
Taron Johnson, Buffalo Bills
Photo: USA Today Sports Images

NFL Week 10 - The Buffalo Bills are back on top of the Football Outsiders DVOA ratings after a big win over the lowly New York Jets this week. The Bills are going to be an excellent test of our precept that championship teams are defined mostly by their dominant wins over bad teams. The Bills are the least consistent team of the year according to week-to-week variance of DVOA. They have four wins with DVOA over 60% but they also have three losses with negative DVOA. Those losses weren't to particularly good teams, either, at least according to DVOA. Yes, the Titans are 8-2, but all three of the teams that beat Buffalo are in the bottom half of our DVOA ratings.

The team Buffalo replaces on the top of the DVOA ratings is the Arizona Cardinals. Obviously, the Cardinals fell out of the top spot in large part because they had to use their backup and third-string quarterbacks against Carolina this week and ended up with a big 34-10 loss. However, it is worth noting that this week was Arizona's worst performance of the year on both offense and defense. Even if Arizona had the same offensive DVOA as a week ago, the Cardinals still would have fallen out of the top spot overall because of the decline of their defense this week. They would just be No. 2 instead of No. 4.

As we often have to point out, the DVOA ratings are not adjusted for backup quarterbacks. DVOA measures what happened on the field based on the players who were on the field in those games, and we don't estimate what teams would have done if they had different personnel on the field. However, we do estimate the effect of backup quarterbacks on predictive ratings going forward in the DAVE ratings that we use for our playoff odds simulation. DAVE ratings combine weighted DVOA with preseason projections, giving more weight to games over the past two months while also incorporating September games as well as prior data from before the season. For example, DAVE assumes that Kansas City is better than what the Chiefs have shown us so far, but each week the amount of preseason projection we use in DAVE goes down as we get more data about how good Kansas City actually is this season.

The DAVE ratings listed below do not adjust for backup quarterbacks, but like I said, we do estimates for the effect of backup quarterbacks when we do the ratings that we use in the playoff odds simulation. Here's a look at the difference right now between the top teams in DAVE and the top teams in quarterback-adjusted DAVE.

Rk Regular
DAVE
QB-Adjusted
DAVE
1 BUF 25.6% BUF 25.6%
2 TB 24.7% TB 24.7%
3 NE 19.3% DAL 22.0%
4 DAL 19.2% ARI 20.0%
5 ARI 17.0% NE 19.3%
6 LAR 14.7% LAR 14.7%
7 SF 14.5% SF 14.5%
8 NO 11.4% SEA 13.5%
9 MIN 11.4% GB 12.7%
10 GB 10.3% MIN 11.4%
11 BAL 10.0% BAL 10.0%
12 KC 8.7% KC 8.7%

Dallas, San Francisco, Minnesota, Kansas City, and Green Bay all moved up in DVOA this week, but the biggest mover was the New England Patriots. The Patriots move all the way up from 13th to fifth in DVOA after a huge dominant win over the Cleveland Browns. Opponent adjustments reached full strength this week, and with the current opponent adjustments, the Patriots get 126.0% DVOA for this win. That's the second-highest single game of the season, trailing only Buffalo's Week 4 40-0 castration of the Houston Texans. That game is currently at 126.7% DVOA. It would be over 150% if not for opponent adjustments! With their rise in both DVOA and DAVE, we now have the Patriots making the postseason 88% of the time in our playoff simulation.

Cleveland, meanwhile, plummets from sixth to 17th in DVOA off this one terrible game. The Browns' playoff odds also drop 24% from 56.5% to 32.5%.

If you look at the playoff odds report, you'll notice that there's a big difference between the most likely teams to make the playoffs and the most likely teams to win the Super Bowl. At this point, the former is based more on the team's record and their schedule going forward. The latter is based partly on what seed a team is likely to win but also significantly on the team's DVOA rating. That's how we get Tennessee and Green Bay, each at 8-2, as the teams most likely to make the playoffs but not in our top five most likely teams to win the Super Bowl.

We've written a lot about Green Bay and Tennessee this year. The Packers moved up in DVOA this week with a nice win against Seattle. Their defense has really turned things around over the last month. Green Bay's defensive DVOA for Weeks 1-6 was 11.4%, which was 27th in the league. Green Bay's defensive DVOA for Weeks 7-10 was -22.3%, which was sixth in the league. The Packers are still being held down in DVOA by their terrible Week 1 loss to New Orleans and by a lower offensive DVOA for the game where Jordan Love played quarterback while Aaron Rodgers was out for COVID.

As for Tennessee, well, we've written a lot about the Titans and they did it again this week. They won another game despite having the lower DVOA rating. Tennessee beat New Orleans despite being outgained significantly, 373 yards (6.1 per play) to 264 yards (4.6 per play). They won the turnover margin because each team fumbled once but the Titans recovered both fumbles. New Orleans' fumble came on a kick return and gave the Titans a short field at the New Orleans 19. They scored a touchdown. And the Titans took advantage of both a bad fourth-and-goal call by Sean Payton, kicking a field goal from the 1, and two missed extra points they really had nothing to do with. And still, the Saints were within a two-point conversion of sending the game to overtime.

The Titans get -19.9% DVOA for this game, which lowers their season overall DVOA to 0.7%. That's still above average, slightly, but it ranks 18th in the league. The Titans are a little higher in weighted DVOA, 3.5%, which ranks 15th. Their very easy remaining schedule and 8-2 record give them a significant leg up on getting the No. 1 seed in the AFC and that all-important first-round bye. Even with such a low DVOA, they currently get the first-round bye in 48% of our simulations. But the Titans simply are not as good as their record and they should not be No. 1 or even in the top three of NFL power rankings around the Internet.

Here's where the Titans are now compared to the worst 8-2 teams in DVOA history. All of these teams made the playoffs. Some of them even made the Super Bowl! But the ones that made the Super Bowl generally improved in the second half of the year. I'm curious to go back and see what the press was saying about some of these teams at the time. Were the 2000 Vikings on top of NFL power rankings around the Internet? I remember having to explain the 2003 Panthers constantly, since that was the first year of Football Outsiders. They actually declined in the second half of the season, then turned things around and got hot in the playoffs to almost win the championship.

Worst 8-2 Teams by DVOA, 1983-2021
Year Team W-L DVOA Rk Final
W-L
Final
DVOA
Final
Rk
2000 MIN 8-2 -3.5% 20 11-5 -1.3% 20
2021 TEN 8-2 0.7% 18 -- -- --
1999 TEN 8-2 4.2% 18 13-3 14.9% 5
2003 CAR 8-2 5.9% 14 11-5 -0.2% 16
2021 GB 8-2 6.8% 12 -- -- --
2001 CHI 8-2 7.3% 11 13-3 15.9% 8
1997 MIN 8-2 7.5% 10 9-7 -0.8% 16
2004 ATL 8-2 7.7% 12 11-5 -1.7% 15
2016 OAK 8-2 8.7% 8 12-4 7.6% 10
1989 NYG 8-2 9.0% 9 12-4 15.3% 6
1998 ATL 8-2 9.3% 11 14-2 22.9% 6
2002 GB 8-2 9.8% 12 12-4 7.8% 13

* * * * *

Football Outsiders playoff odds, snap counts, and the FO+ database are now all updated through Week 10. A reminder that all our free stats pages, including DVOA and player position stats, require registration to view. This is not a paywall! You only need to register (for free) and then log in to the site to view these pages. While you're at it, you can get a seven-day trial of FO+ and check out the FO+ features like a deeper DVOA database, weekly fantasy projections, fantasy football research tools, and picks against the spread.

* * * * *

Here is the Football Outsiders Top 16 through 10 weeks of 2021, measured by our proprietary Defense-adjusted Value Over Average (DVOA) system that breaks down every single play and compares a team's performance to the league average based on situation in order to determine value over average. (Explained further here.)

OFFENSE and DEFENSE DVOA are adjusted for opponent and performance indoors and consider all fumbles, kept or lost, as equal value. SPECIAL TEAMS DVOA is adjusted for type of stadium (warm, cold, dome, Denver) and week of season. As always, positive numbers represent more points so DEFENSE is better when it is NEGATIVE.

Opponent adjustments are now at full strength as we have hit Week 10.

DAVE is a formula which combines our preseason forecast with current DVOA to get a more accurate projection of how a team will play the rest of the season. DAVE is currently 36% preseason forecast and 64% actual performance for teams with nine games played, and 24% preseason forecast and 76% actual performance for teams with 10 games played. It is not currently adjusted for any backup quarterbacks.

To save people some time, please use the following format for all complaints:

<team> is clearly ranked <too high/too low> because <reason unrelated to DVOA>. <subjective ranking system> is way better than this. <unrelated team-supporting or -denigrating comment, preferably with poor spelling and/or chat-acceptable spelling>

RK TEAM TOTAL
DVOA
LAST
WEEK
TOTAL
DAVE
RANK W-L OFF.
DVOA
OFF.
RANK
DEF.
DVOA
DEF.
RANK
S.T.
DVOA
S.T.
RANK
1 BUF 33.1% 4 25.6% 1 6-3 4.7% 14 -25.3% 1 3.1% 3
2 TB 28.1% 2 24.7% 2 6-3 24.6% 1 -4.5% 9 -1.1% 23
3 DAL 27.5% 5 19.2% 4 7-2 14.6% 4 -13.7% 4 -0.8% 22
4 ARI 24.9% 1 17.0% 5 8-2 7.5% 11 -14.9% 2 2.5% 6
5 NE 20.1% 13 19.3% 3 6-4 4.2% 16 -13.6% 5 2.3% 8
6 LAR 18.6% 3 14.7% 6 7-3 15.2% 2 -5.9% 7 -2.5% 27
7 SF 17.2% 9 14.5% 7 4-5 15.0% 3 -0.6% 13 1.6% 10
8 NO 16.2% 8 11.4% 8 5-4 4.5% 15 -10.7% 6 0.9% 16
9 MIN 13.6% 11 11.4% 9 4-5 6.5% 12 -5.5% 8 1.6% 11
10 IND 8.2% 10 7.7% 13 5-5 4.0% 17 -3.3% 10 0.9% 17
11 BAL 7.3% 7 10.0% 11 6-3 8.0% 10 5.2% 25 4.5% 2
12 GB 6.8% 15 10.3% 10 8-2 9.3% 7 -2.2% 11 -4.7% 30
13 SEA 6.2% 12 5.6% 14 3-6 8.1% 9 3.2% 21 1.3% 12
14 KC 4.8% 17 8.7% 12 6-4 11.3% 5 11.1% 30 4.6% 1
15 PHI 4.7% 18 1.2% 16 4-6 8.5% 8 3.0% 19 -0.8% 21
16 LAC 2.8% 16 0.5% 17 5-4 11.0% 6 2.2% 18 -6.0% 31

Click here for the full table.

Comments

98 comments, Last at 19 Nov 2021, 9:28am

1 This season is truly bizarre…

This season is truly bizarre in terms of being able to evaluate the top teams in the league. It seems like every time a team ascends to Super Bowl contender status, they suffer an inexplicable and embarrassing loss (think Buffalo vs. Jacksonville, Arizona vs. Carolina, Dallas vs. Denver, TB vs. Washington, LAR the last two weeks). The playoffs should certainly be fun an unpredictable. 

19 Looking more broadly at the NFL...

...22 of the 32 teams in the league, 11 in each conference, have a winning percentage above .300 and below .700.  (That excludes the 7-3 Rams and 3-7 Dolphins.)  Three entire divisions (NFC South, AFC North, AFC West) are included in that 22 team grouping of relative parity.  Only one team in the NFC is worse than 3-6, and only one team in the AFC is better than 6-3.

29 The season probably feels…

The season probably feels more chaotic just because so many of the wild results have come very recently. I commented on the open thread on Sunday that all seven of the teams in the top seven of DVOA going into the week have had an awful loss in the past two weeks. With SF entering the top seven, that's still almost true. The only exception is NE. By the end of the season things might not feel nearly so chaotic and bizarre.

89 TL;DR up front I will admit…

TL;DR up front
I will admit I basically draw no solid conclusions. The stats are a bit fun potentially. My premise of looking for other years where every team had at least 2 losses through the first 10 games may not be the right starting place. The parity this year, in the AFC especially with 2 divisions (West and North) that don't have a team with a losing record could be a bigger factor and better parallel, but it was the 2 loss thing that started my rummage in the stats bin.

Some potentially fun looks back at weird years
I read somewhere that this is only the 4th time since 1970 that every team has at least 2 losses through 10 weeks. So it certainly does seem like a pretty weird year and I wanted to see if we could find out if DVOA did still do the job well, my hypothesis is that it did. So I made a quick script to generate the URLS and flipped through the pfr records as of page to find the other 3. Only 1 was in the DVOA era since it looks like they were 1979, 1981, and 2010.

Since FO was still using formulaic URLs that make finding older stuff easier we can pull up the 2010 Week 10 DVOA numbers. It was a season near and dear to me of course since the after 10 weeks 5th ranked 6-3 Packers did win it all that year. Of course there were no 8-2 teams since every team had had a bye week already that season

So you've got 6-3 Chicago (-13.6%) and 6-3 TB (-9.3%) with negative DVOA's. Those are closest to 5-3-1 PIT (-7.1%) I guess.
7-2 NYJ was the lowest ranked of the 2 loss teams at 11.9% so still higher than this years 8-2 teams TEN (0.7%) and GB (6.8%).

Top 10 + others who made playoffs were:

1. PHI (6-3) @ 34.4%. Finished 4th going 10-6 @ 22.2%. Lost to SB Champ GB in Wild Card
2. TEN (5-4) @ 26.6%. Finished 11th going 6-10 @ 6.0%. Did not make playoffs
3. PIT (6-3) @ 24.8%. Finished 2nd going 12-4 @ 34.7%. Lost in SB to GB
4. NE (7-2) @ 23.6%. Finished 1st going 14-2 @ 44.3%. Lost to NYJ in Divisional Round
5. GB (6-3) @ 22.9%. Finished 3rd going 10-6 @ 23.5%. Won SB
6. NYG (6-3) @ 21.4%. Finished 10th going 10-6 @ 11.8%. Did not make playoffs
7. SD (4-5) @ 17.2%. Finished 8th going 9-7 @ 14.7%. Did not make playoffs
8. BAL (6-3) @ 16.5%. Finished 5th going 12-4 @ 21.9%. Lost to PIT in Divisional Round
9. IND (6-3) @ 16.0%. Finished 17th going 10-6 @ 1.0%. Lost to NYJ in Wild Card Round
10. ATL (7-2) @ 13.0%. Finished 6th going 13-3 @ 20.4%. Lost to SB Champ in Divisional Round
11. MIA (5-4) @ 12.7%. Finished 16th going 7-9 @ 1.6%. Did not make playoffs
12. NYJ (7-2) @ 11.0%. Finsished 7th going 11-5 @ 17.7%. Lost to PIT in AFCCG
13. NO (6-3) @ 7.8%. Finished 9th going 11-5 @ 12.5%. Lost to SEA in Wild Card
15. KC (5-4) @ 6.4%. Finished 18th going 10-6 @ -0.9%. Lost to BAL in Wild Card
24. CHI (6-3) @ -12.4%. Finished 14th going 11-5 @ 2.2%. Lost to GB in NFCCG
29. SEA (5-4) @ -22.1%. Finished 30th going 7-9 @ -24.3%. Lost to CHI in Divisional Round

The AFC seemed to mostly sort itself out as the season finished. NE didn't lose again and PIT and BAL each lost 1 more time. They looked quite strong at the end and the SB rep for the AFC came from that group. The NFC though... That was the 7-9 division winning Seahawks. TB, GB, and NYG all finishing at 10-6 and GB getting in through tie breakers (CHI won the division at 11-5) and 2 10-6 teams from the same conference missing. So yeah the all having 2 losses did seem to predict some playoff weirdness.

What I don't have is the data for how frequently the top 10 DVOA teams after week 10 make the playoffs. DVOA in 2010, the last weird year with all teams having at least 2 losses after week 10 looks like it did alright. TEN bombed after getting to 5-4 with a good DVOA against the toughest schedule (they are mentioned in the article). NYG didn't make it, but they did still go 10-6 and missed from tie breakers. SD was edged out by KC in the division but DVOA isn't the only metric that thought SD was better. Also under the current format SD would have made it in.

Seems like 2010 had more to do with unbalanced schedules thanks to a really crappy divisiion
SEA already mentioned. They won a trainwreck division, that was awful all season (St. Louis was leading it at 4-5 after week 10). Someone was going to win it. SEA pulled the upset over NO in the playoffs and then lost to the other low DVOA team, CHI. So CHI going to the NFCCG had a lot of luck that was in part driven by the NFC West being so bad. So maybe DVOA wasn't too off about them either since all the "bad" NFC teams were in the same playoff bracket. The 2nd best NFC team at the time (GB), ended up ranking slightly ahead of the best team at the time (PHI) but because of the weird records they squared off first game and GB won, so a feather in the DVOA cap there. Then the 4th best NFC team as of week 10, but 3rd best NFC team by the end lost to GB in the next round. So yeah the top 3 NFC teams were all on the same side and the bottom 3 were on the other side. Again in part because the NFC south played the NFC west and inflated records helping them get one of the wild cards and shuffling the seeding.

But this is a very narrow look and with TEN and CHI there are couple of strong points against DVOA. The other misses as mentioned aren't too weird. I just wondered if it could give us any help with the oddities that are 2021. DVOA seemed more right than W-L records in 2010, but the the W-L records were also closer to what DVOA said. The NFC had really weird records because of the NFC West dumpster fire. There isn't a division that bad this year.

Schedule strength effects are there any?
So will SF with their 17.2% DVOA and 4-5 record against the 5th toughest schedule crater like 2010 TEN with their 26.6% DVOA and 5-4 record against the toughest DVOA schedule did? Is a good DVOA and middling record against a tough schedule non predictive?  2021 has 2 more teams with winning records against tough schedules. LAC toughest schedule, but only 5-4 and 2.8% DVOA so that doesn't feel odd. TEN 8-2 with a 0.7% DVOA against the 2nd toughest schedule. Does it mean anything?

 

2 Outlier Games

It seems like outlier games have a huge affect on teams in the rankings, especially looking at Green Bay. Its interesting 2 games disproportionately holding down a team, one where the starting QB didn't play and the other the first game of the season. If a team was to beat GB right now, they would also not get as much credit for the win due to those losses making GB seem comparatively weaker.

20 Outlier Games

In reply to by mkuch90

It isn't just GB. Pretty much every team at the top is there due on the back of a few Gigantic Wins. The Bills, Bucs, and Cowboys all have had some crushing victories.

Look at NE, they went from 13th to 5th after a 45-7 Win. I'd bet the 45-7 Win and their 54-13 Win make up a pretty large percentage of their ranking. Does NE's annihilation of shaky QB's translate to predictive playoff success without also having a Hall of Fame QB? We'll see.  

GB by contrast doesn't have a win over 18 this year.  If I had to bet, I would bet they would not win any game this year by 20. They'll never get the +120% DVOA boost to skyrocket them in DVOA. We'll see how it plays out.

27 It's one of the ideas of…

In reply to by Q

It's one of the ideas of this site that blowout wins tell us more about a team then tight victories. 'Stomps' are more predictive than 'guts.' There is distinctly nothing surprising about teams being ranked high on the back of a few big victories.

35 Q

Yes, but there is also more nuance than just big win > close win.

Is a 50 point win really more predictive than a 30 or 40 point victory? Is a 200 point victory that much more predictive than a 70 point victory? Theoretically, a team could be 1-15 and #1 in DVOA entirely based on the strength of 1 absurdly, impossibly strong performance (even if it had a negative DVOA in all other 16 games).

It also is hard to deny that teams to varying degrees have different philosophies on running up the score, pace of play, etc.

49 They'd need a single game…

They'd need a single game DVOA in the neighborhood of 350-400%. Which I think is probably technically mathematically possible but obviously would require something incredible. Even COVID-affected games last year weren't even remotely close to that.

82 I do think there's…

In reply to by Q

I do think there's diminishing returns in the magnitude of blowouts. I've always thought season DVOA would be more accurate if single games were capped at like 75%. 

98 DVOA isn't averaged by game,…

DVOA isn't averaged by game, it's averaged by play. There's no way to cap a game at a value. The only real thing you could do is essentially discount the plays late in a blowout, but the outliers in a blowout aren't the late game plays. They're the early game plays that result in the blowout. When teams scream out to a big lead in the first half, they almost always slow down in the second half, because they switch to running the ball to burn out the clock.

The problem with reducing the importance of blowouts is that it's hindsight - you effectively want to discount the huge lead because of the end result.

Take the Patriots/Jets game, for instance. If you say "we're gonna throw out all the plays in the second half, they were already up 24 points at the half." That makes the problem worse, not better.

53 NE basically has 4 really…

In reply to by Q

NE basically has 4 really good DVOA games - killing the jets and Cleveland, and basically tying Tampa and Dallas, and two really bad games (NO, and Houston) 

63 NE is clearly ranked too…

NE is clearly ranked too accurately because they just barely lost to TB and DAL, and are ranked just below them in DVOA.  DVOA is more complicated than just looking at blowouts.  Speaking of blowouts, there's probably a "yo mama" joke to be made here.

3 Parity

Everybody is clearly ranked too high because there are no truly good teams this season.  Raiderjoe's spelling is way better than this.  The FOMBC is the only successful aspect of this season, as it seems universally applicable, even if unrelated to the actual FOMB.

5 Well done

In reply to by young curmudgeon

Bravo. It's true which makes it funnier. 

6 Welcome Titans to the Club

I have decided to name a certain aspect of DVOA a "Packer"

This where Team A beats Team B, often on the road; Team A has a better record than Team B, often by several games... and DVOA stubbornly insists that Team B is superior.

Should GB win narrowly at MINN this weekend, that could give the originators of the "Packer", the Packers, 3 such "prizes" for 2021. Wins at Arizona, SF and Minn-- all of which have worse records (OK, the Cardinals are because of the tiebreaker, but technically still true) all rated more highly by DVOA...  The Rams game after the BYE at Lambeau could be Numero 4.

But this year GB are pikers compared to the true champions of the "Packer". Tennessee now can claim an astounding 7 of them-- nearly their entire win total for the season.

The Titans have defeated IND (twice!), KC, BUFF, SEA, RAMS and NO and are a combined 18 games better in W/L than those 7 opponents-- yet DVOA has them all rated better-- and by a whopping average of 10.6 positions... (Colts are counted twice because... well, you see the Titans beat them in both games)

So 7 Packers for the Titans-- I bow in admiration.

Another more sympathetic fun fact to the difficulty of any metric figuring out this season:

After 8 weeks, the supposed Big Six of the NFC-- the 4 division leaders plus LAR and NO-- were 38-8 combined. In the last two weeks, those six teams are 3-8. Not a misprint.  Nobody knows anything...

9 the entire point of DVOA

The entire point of DVOA is that W-L is less predictive than other stats.  The Packers have the best record in the NFC, but the Cardinals, Rams, Bucs, Saints, and Cowboys all have better point differentials.  Some of them have played 1 fewer game. And the Packers apparently have terrible special teams weighing them down. 

As for all the upsets: we've hit a part of the midseason when teams are finding it hard to keep focus every single week, and that seems to leave them open to upsets, esp. if they're missing a key player or two.

18 DVOA

W-L is definitely not the best way to evaluate teams. It is also worth noting that teams care differently about margin of victory. If you have 2 identical teams and 1 team goes into run out clock mode up by 10 late in the game whereas Team B goes into try to win by 24 mode late in the game, DVOA will favor Team B since it appears to B the more dominant team since it won by 24.

I of course have no way to deal with this and know that in general point differential is 1 of the more predictive methods. I also know that some teams care more about turning 14 point leads into 30 point leads than do other teams.

If my money was on the line, I would bet on a Buff/TB Super Bowl. However, I would also favor GB over any other NON TB NFC Team.

28 Those differences in caring…

In reply to by Q

Those differences in caring about how large the final scoreline is are marginal and are not going to make hardly any difference in point difference. Even the best teams spend very little time overall in a whole season far enough ahead they are worrying only about how much they will win by. Whatever strategy differences there are will make virtually no difference.

36 Q

I think your assumption is flat out wrong. If a team is up by 17, late into the 3rd, trying to run up the score vs taking the foot off the gas could easily be the difference bewteen a 7-10 point win and a 28-31 point win.

In that game alone. That could be a 21 point difference. Over the course of a season that could be a 50-100 point swing.

If you think +/- 50-100 points doesn't radically swing how a team will look to analytical models then we have vastly different beliefs.

54 Sure .   But those teams…

In reply to by Q

Sure .

 

But those teams that keep piling up points win more.

 

​​​​Both in that they don't blow leads as often, and that continuing to be able to put up points is predictive

55 Q

You are partially right and partially wrong. At the cumulative level it may be predictive but that doesn’t make it predictive at the individual level.

Additionally, playing to maximize your points by playing at a faster pace, keeping your starters in longer, continuing to throw, etc increase your injury odds which are all very negative long term effects.

 

62 Brady’s Patriots never had a…

In reply to by Q

Brady’s Patriots never had a problem doing the things that you just mentioned.  Their slaughter victories proved to be very predictive of success.

70 Again, there's 30 years of…

In reply to by Q

Again, there's 30 years of data compiled on this very site showing that these things are highly predictive. 

 

 

 

71 Congratulations, you've…

In reply to by Q

Congratulations, you've learned the difference between having many data points vs. only one.

64 You seem oddly confident…

In reply to by Q

You seem oddly confident about this, considering that you - by your own admission - do not have the ability to test the hypothesis and verify it. 

16 I think you’re just arguing…

I think you’re just arguing that randomness exists. That’s true. 

DVOA has a pretty good track record once you have a large enough sample size. It’s not meant to be taken without consideration of other factors, nor is any statistic in any realm. It doesn’t account for positional strengths/weaknesses that could become critically important in a particular game, nor does it understand a teams ability to play in weather, nor any other random item that is also important to predicting games. It does not give credit for watching team after team implode in your general direction. 

Maybe the Titans will be good by the end - seems a little unlikely so far. 

8 TN in the bottom half?

Don't see that changing much because they're gonna slide to the #1 seed with the easiest SOS remaining. 

Packers D up to 11. Looks good without Jaire and Zada so we'll see how Joe integrates them back in at the end of the season.

11 DVOA and the 4th down

A number of your fan contributors were commenting in this week's open discussion about DVOA and 4th down.  Then one of your writers wrote this article:

Frank Frigo

in

Risky Business

The Year of the Fourth Down Revolution?

Although Frank admits we have a long way to go, it is clear that going for fourth down is increasing.  

 

Aaron, we had discussed this two years ago when on 3rd and 17 against Seattle Lamar Jackson gained 15 yards leaving 4th and 2  Baltimore goes for it, and Baltimore scored a TD.  You said that the 3rd down play gets partial credit.

Now this week, we had these two huge plays:  MIN on 3rd and 20 gains 18 yards on pass to Thielen then converts 4th and 2, and Goff on 3rd and 6 completes a 5 yard pass (otherwise known as Goff airing it out) and then DET converted 4th and 1.

Two questions:

1.  Do you feel that going forward that since more teams are playing with 4 downs that the DVOA formula will change?  I see the counter to this which is that of course being a yard or two short on a long 3rd down play could lead to a catastrophic failure on 4th down. (I mean just getting stopped in poor field position, although Fangio may never go for 4th down again).

2.  I believe that you had mentioned a few weeks ago that DVOA may need to add in coaching decisions.  We are in a league where some teams are regularly playing with 4 downs and some are not.  Ironically some of the all time most successful coaches are so conservative and are not with the analytic age (Belichick, Tomlin, Punting Pete Carroll) and others are all in on 4th down (Harbaugh, Staley, Campbell).  Do you think that it is time that to accurately analyze team performance, that coaching decisions must be included?

50 Re-evaluating DVOA for 3rd…

Re-evaluating DVOA for 3rd/4th down.

I think if you want to credit a 3rd down that’s not converted, then it should require going for it on 4th down the next play and that it becomes 4th&3 or less (or some very high % of the yards to go from 3rd down).

68 I think that you need to be…

I think that you need to be careful with regards to saying that the third down play is a success only if the team goes for 4th down with a high chance of succeeding.  

DVOA is a measure of what happened.  The 3rd and 20 play in which Thielen gained 18 yards needs to stand on its own, just like a first down sack of -7 needs to stand on its own, regardless as to whether the team converted 4th and 2 or 2nd and 17.  The sack is a bad play regardless as to whether the team dug itself out of the hole. The value of the 18 yard play is subject to debate, but I believe that for consistency of the DVOA formula, each play must stand on its own.  

The game value of all plays in football are contingent upon other plays that occur. This is not true for DVOA.  After a kickoff touchback, a 74 yard pass to the 1 is a great play regardless as to whether a lost fumble occurs on the next play.  

One concept that is used in baseball that I believe should be used in football is the concept of "clusterluck."  In baseball if a team gets 2 walks and 7 singles as their entire production for the game, this is bad offense.  However, if they get 7, 8, or all 9 of these baserunners in the same inning, they will score an above average amount of runs.

Similarly in football, if a team gets most of its effective plays on 3-4 drives, they may score 21-28 points despite having a poor offensive outing.  Basically if this theoretical game occurred and the team scored these points, they may have done so with less than 300 yards of offense, which is a poor game of offense, despite having an above average offensive point production total.

86 Similarly in football, if a…

Similarly in football, if a team gets most of its effective plays on 3-4 drives, they may score 21-28 points despite having a poor offensive outing.  Basically if this theoretical game occurred and the team scored these points, they may have done so with less than 300 yards of offense, which is a poor game of offense, despite having an above average offensive point production total.

 

Check out Buffalo @ Miami, from week 2 of this season. 

12 DVOA Less Predictive in 2021

When I look at the large variances associated with the top rated teams I see 8 of the top 16 teams on DVOA have variances in the top half of the league.  If I understand this correctly, this means that the week-to-week performance of a top team on the DVOA metric is highly variable.  In other words, the league is more of a matchup-based league with top teams losing to weaker teams in 2021.  Is this the right way to think about it?  If it is, do we think the league has adjusted to recent new QB talent with scheme and this is allowing weaker teams to game plan to defeat top teams on any given Sunday?

25 Maybe

But I think it’s more likely, or at least as likely, that some of the high-variance teams have crushed teams so thoroughly that their middle-of-the-road games and outright bad games look much worse by comparison.

78 You Can Take This Fact Either Way

W/L means nothing... or W/L means everything (to a coach's future, we know which one is more true)

Here are the next game records since Week 4 of the 8 DVOA quadrants. Just the next game, because after that the metric shuffles and changes the rankings. And yes of course where those games were played, against whom, and affected by which other factors (Injuries/COVID, etc.) can affect these outcomes-- but those same factors also affect DVOA, too.

Teams Ranked #1-4 in DVOA-- 13-9  the next week

#5-8         12-9

#9-12        9-12

#13-16      16-7 (!)

#17-20      10-11-1

#21-24      10-12

#25-28      9-13

#29-32      6-12-1

So if I can cherry pick-- teams ranked #1-12 are 34-30 the following week, barely over .500  Teams ranked #13-28 are 45-43-1 the next week, virtually the identical percentage. That's all the supposed cream vs. nearly all the mediocre to bad, excluding the very, very bad. And there's no difference. of course the outlier performance by the #13-16 is what creates this-- I said I was cherry-picking...

Point being-- this is a wacky league. No metric predicts particularly well.

The 16-7 record in that #13-16 sweet spot comes mainly from the following teams:

NE (3 wins)  finally this week the Patriots jumped to where they deserve to be

MINN (2 wins)

GB (2 wins)  Vikes v Pack in the underappreciated bowl Sunday

The Titans only have one win in this company-- their others came when ranked 30, 27, 27, 26, 23, 20 and 20

13 DVOA Adjustments

I wonder if throwing out the best and worst game for each team would cause better correlation of DVOA. I imagine that wouldn’t work early in the year, but wonder if it’s advantageous by the time you reach midway through. Has anything like that been attempted? 

33 Sample sizes are so tiny in…

In reply to by Wifan6562

Sample sizes are so tiny in NFL seasons, I would suspect that making the samples even smaller would subtract information rather than improving projections.

I know that's not a direct answer to your question.  Just my a priori expectation.

14 Buffalo's defense is over 10…

Buffalo's defense is over 10 points of DVOA better than the next nearest team. That's crazy. The game against Carolina should be a brutal defensive struggle.

21 Buff's D

We'll eventually see how real Buff's D really is. By my count they have faced 2 good offenses this year. Tenn scored 34 which isn't anything to write home about. KC scored 20 although they have been very up and down so hard to evaluate.

In some ways they remind me of the Pats from Brady's last year (Must be nice to play in the AFC East!!!).  Through like 1/2 or 2/3 of the year they graded out like the 85 Bears since it was a never ending line of High School Caliber QBs that they absolutely obliterated.

Here are the powerhouse other offenses they have faced this year: Steelers, Dolphins, Redskins, Texans, Dolphins Again, Jaguars, Jets.

I know DVOA adjusts for opponent adjustments, but some teams/units particularly destroy awful opposition.

23 DVOA takes time

In reply to by Q

Buffalo's had an incredibly light schedule.  And they've been beating up on weaker teams.  Which is fair.  But I think they're getting way too much credit for beating Miami 35-0 and Houston 40-0.  

Well, things are going to get a bit harder.  So far they've only beaten one team with a winning record: KC.  They're getting Indy, NO, the Pats, Bucs, Panthers, and Pats again in their next six games. That should drag their DVOA down a bit.

34 The FO guys have been…

In reply to by RickD

The FO guys have been consistent, however, in confirming that those 35-0 and 40-0 results are predictive of future success.  While we (humans) tend to think "yeah, but they only beat HOU, everyone beats HOU", past history suggests that beating HOU 40-0 says a lot about the likelihood of BUF prevailing in future games against better opponents.

As you say, their next few games will tell us a lot about this BUF team and whether they're as good as DVOA thinks they may be.

38 Like any factor model, you…

Like any factor model, you have to evaluate the specifics and see if there's any logic to why it might explain it, or if it's just an unusual correlation. The way I always thought about it, Stomps are more predictive than Guts because there typically isn't a huge gap in talent between, say, the #1 and #20 team in the league, so that bad teams almost always have a puncher's chance at a win even against contenders. Guts are not predictive because if most teams trend towards the median, most games should be close wins or close losses. Stomps are predictive because they only happen if one team is significantly better than the other.

Where I think it might fall down a bit is that the salary cap means there's an upper bound to just how good a team can be, but even with a salary minimum, there's no lower bound on how bad a team can be. We've seen multiple teams over the last decade where the GM was openly tanking to rebuild from scratch, despite the best efforts of the coaches and players. And Adam Gase. So even if the league generally converges around the median, the very worst teams are so bad that stomping them might not be as predictive as an ordinary 4-13 bad team.

I have no idea how this applies to Buffalo, except that losing to Jacksonville makes my head hurt.

95 @IndependentGeorge: Stomps…

@IndependentGeorge: Stomps are more predictive than Guts in large part because Gutty wins are more greatly affected by random elements -- weather, fumble bounces, injuries, bad referee calls, etc. The ball's not round, after all.

45 Buffalo vs Brady's Patriots

The Patriots would destroy teams in the prime of Brady's career.  NE often had a cakewalk division and would often destroy opponents.  Yes, they played the first place schedule most years, but division winners would often move backwards, with their division not moving forward.  Thus, NE often had very easy schedules. 

What makes 2021 Buffalo different from the Brady NE days is that Buffalo is doing the annihilation mostly on the defense end.  We know that defense is less predictable than offense.  We always know that statistically that the absolute top and absolute bottom should regress towards the mean as the season progresses.  

Thus, no matter who Buffalo plays going forward, we would expect a regression towards the mean.  When regressing from 33.1% we will find exactly what is playing out, that there are no exceptional teams this year, resulting in many upsets and that both conferences are wide open as to the winner, and eventual Super Bowl entrants. 

On the flip side there are no exceptional awful teams this year although I would have trouble trying to convince the Lions fans of this, despite the fact that they are not last in DVOA.  We are not too far removed from 2019 Miami, who with a tremendous second half run ended up at -38.8 DVOA.

57 Q

Not really. It could be very matchup dependent like in a Video Game.

If you show up against a Boss that has only Ice Attacks and you have a 100% Ice Immunity Ability, against that Boss you are going to look like the most Overpowered Being in Existence.

If the game takes place in an Ice World then you are in great shape and that matchup will likely be quite predictive of Future Encounters. However, if that is the only Ice Level out of 50 and the other 49 are in the scorching desert against Fire Monsters then its predictive value becomes essentially 0.

67 Let me see if I've got this…

In reply to by Q

Let me see if I've got this right: if NE goes to Lambeau they look unbeatable, but the Super Bowl is always* played in a warm-weather city or dome, so they'll never win a Super Bowl, right?

* okay, now almost always

</sarcasm>

72 More like, the Bills just…

More like, the Bills just trainwrecked a guy with questionable arm strength who can be accurate with good timing on short passes.  The only team that did well against them offensively was Tennessee, and that was because King Henry ran over them.  Maybe the Patriots can run on Buffalo enough, but if not, it's all on Mac Jones, and that's not a good plan.

The Colts have Taylor, so that's a rough matchup for the Bills defense.  We'll see what happens.

77 More like, the Bills just…

More like, the Bills just trainwrecked a guy with questionable arm strength who can be accurate with good timing on short passes.

I'm sorry, my brain's just refusing to comprehend this, I guess. Are you really suggesting that how the Bills did against the Jets might foretell how they'll do against the Patriots? The Jets game is what you're referring to, right?

I just... I mean, I get that you could make some sort of vague comparison between White and Jones or something (serious stretch, but whatever) - but, well, the Jets are... "schematically limited." Let's put it that way.

31 The Jets had 900 yards of…

In reply to by Q

The Jets had 900 yards of offense the two games before they played Buffalo.  Mike White had 116 DYAR against the Bengals and 43 DYAR in two drives against Indy.  Josh Johnson had 32 DYAR against Indy.  That offense was on a roll, and the Bills stuffed Mike White and co. into the wood chipper.  You can hope that doesn't happen to Mac Jones, but I'm kind of doubting it.

 

52 I think those are just…

I think those are just matchup differences. Buffalo reminds me of the "No Fly Zone" Broncos in that all of the DBs are playing out of their minds this year (not that they aren't good, but this isn't really sustainable across a career, IMO, especially from Wallace.) This was actually mentioned in either Tanier's column or Audibles - the Bengals and Colts do not have good pass defenses. The Bills definitely do not rely on their pass rush, because that's pretty damn average.

 

I expect we'll learn a few things about each team this weekend when they play the Colts. If Wentz makes questionable throws in this game, they will bite him. If Buffalo falls into the "run the ball every 3 downs" Daboll-ism, the offense will struggle. The Colts can run the ball and the Bills can be run on, especially with Lotulelei out.

Should be fun! 

73 Those are match-up…

Those are match-up differences, but the Jets offense was also terrible against Atlanta (31st in pass defense DVOA), so that offense definitely improved those two weeks until running into the Bills.

I agree that Colts are a tough matchup for the Bills defensively.  Allen should torch them though.

79 No, they didn't "definitely…

No, they didn't "definitely improve" 

 

 

Variance is a thing, and matchups are a thing. Having a good or bad day is a thing.  

 

 

 

There's absolutely no reason you should be assuming some sort of meaning ful trend out of 2 games, or assuming some other value from a regression to the mean. 

 

 

47 Castration

Well, I suppose that's a steer on their helmet and not a bull.

In that case, quite fitting.

And they probably gave them an ice pack afterwards, to reduce pain and swelling and soften the blow, as it were.

22 It blows my mind to see SF…

It blows my mind to see SF rated #3 on Offense given how awful they look a large percentage of the time.

Their defense is even more confusing as they played poorly against the Bears and 2nd string Cardinals and then very well against the Rams. It makes no sense at all.

44 Kyle Shanahan has started to…

Kyle Shanahan has started to get some major scrutiny this year, even from some of the more measured members of the football media - and he deserves it, to be sure - but the dude can coach offense. He has Garoppolo back up at 8.3 Y/A this year while he's 22nd in air yards/attempt. With Kittle missing significant time and a mostly crap WR corps, too.

I'm a little more surprised that SF's defense has settled in around league average, but they obviously had a hell of a game vs the Rams.

24 Titans

The Titans are the 18th best team in the DVOA and the second worst 8-2 team since 1983. That’s somehow the case even though they’ve beaten the Bills (#1), Rams (6) and Saints (8), and Colts (10), and Chiefs (14). 

40 A team that gained three…

In reply to by EasyLikeSunday…

A team that gained three yards every play, without fail, would have both a terrible DVOA and also go undefeated.  Extremely consistent below-average play can produce above-average results.

42 This. Real world example:…

This.

Real world example: several Patriots playoff victories (most memorably 28-3) of the past decade.  Too lazy to check the exact stats, but while their yds/play were significantly worse, they had waaay more plays.

76 28-3

Pats ran 93 plays, avg 5.9 Y/P while ATL had only 46, avg 7.5 Y/P.  NE was 8/15 on 3rd/4th while ATL was 1/8 and other than needing the one 4th down conversion were perfect during the rally.  Starting at 6:29 of Q3, the kickoff following ATL's final score, NE ran 41 plays to ATL's 16 in regulation and 8 more for the win.  Falcons got some pressure on NE's first couple plays in their drive to tie but ATL's pass rush then disappeared.  About as gassed as possible.

74 They would probably have a…

They would probably have a very high Offensive DVOA because they wouldn't turn the ball over, ever. Whether they would go undefeated would depend upon the Defense (are they allowing 3 yards a play without turning the ball over?), and coaching.  If their coach decided to punt on every 4th and 1 in their end of the field, they could end up losing a lot of games.  Also, if their defense gained 3 yards on every play (either a run sack or a real sack), they would go undefeated and have an enormous DVOA.

26 Situational DVOA

Another theory would be that teams have adapted their game planning to the elite teams (KC, Tampa, Buffalo, LA Rams, Baltimore etc.) and these teams are dropping games they would have won over the past few years.  Not all teams are capable of beating every elite team but some teams can beat one or two of them and this is causing elite teams to display a higher degree of variance from week to week.  Obviously the presence of truly terrible teams on the schedule also creates higher variance too.  Looking at KC's season so far tends to fit this as their offensive woes disappear as soon as they play Las Vegas which uses a single high safety on defense.

30 The 2000 Vikings had the…

The 2000 Vikings had the flashy offense of moss Carter smith and culpepper. They had been to the playoffs four straight years so there was certainly no talk that they weren’t as good as their 8-2 record. That talk was reserved for my Jets who were 6-1 but had all sorts of comebacks. 

also one of the 2 in minny’s 8-2 was the famous “he did what” Antonio freeman Monday night game. 

43 Rams in Trouble?

Of the top NFC teams, the Rams appear to have suffered the most concerning losses so far.

The Packers' losses were the result of having to play a backup QB and what seems to have been very correctable beginning-of-the-season unpreparedness, and it's not like getting blown out in the regular season has prevented them from going to the championship game for the last two years, although that might be their ceiling yet again this year. 

The Cowboys lost a close game to the Super Bowl champs and then appear to have suffered an overconfidence induced, Texas sized stroke while starting a rusty QB against a subpar opponent that's unlikely to be a factor when they're facing top tier opponents.

The Cardinals lost a close game to a good Packers team and then another with a backup QB, so hardly anything to complain about there.

The Buccaneer's losses should be a little more worrisome for them, because they've all been by 9-10 points, although the loss to Washington seems like a case of the no name team simply wanting it more, a revenge game unlikely to play out the same way twice. It's also hard to imagine that any of this will actually matter come playoff time when Brady takes his dose of Tampa Tom Oxy.

The Rams, on the other hand, have had losses of 17, 12, and 21 points, losses that didn't appear all that competitive. Matt Stafford seems intent on turning the ball over, and the latest additions of Von Miller and OBD almost somehow feel more like desperation or a marketing ploy. Not that they won't make the playoffs, but it's getting harder to see them winning their first game with the bad performances they've been putting out.

 

46 Patriots Pass DVOA

Mac Jones Pass DVOA:-1.5%. He has a small DYAR: 216. Essentially, pretty much in the middle of the pack, considering the long tail of some truly wretched quarterback play that populates the passers who don't have enough attempts to qualify for the main table.

New England Patriots Pass DVOA: 18.5%

Can the Patriots highly successful trick plays and Brian Hoyer's successful mop-up drives create such vast separation between the primary quarterback's DVOA and the team's DVOA? 

Is there some other factor included in team DVOA that doesn't get counted for a quarterback's or vice versa?

It just seems to be such an extreme difference, when Jones's sample size is much larger than the other throwers. 

65 A good question!

This is a good question about QB DVOA vs team DVOA. The difference is what goes into the baseline. 
 

QB pass DVOA is only compared to other pass plays. Team DVOA - either pass or rush - is compared to ALL plays. That’s why most teams are above average (0%) passing and below average rushing. 
 

New England’s RANK in team pass DVOA should be much closer to Jones’ rank in QB DVOA. 
 

Hope that helps!

83 Thanks Aaron

In reply to by Aaron Schatz

That certainly does explain the difference. I appreciate the comparisons much better now. Thanks,

48 Colts D and garbage time in DVOA calculations

I'm still boggled by Indy's overall D rating.  I've said it before:  up 17-0 in the first Q, it becomes garbage time for Indy (last week, the WHOLE TEAM failed to show up after that, and after Taylor had like 94 yards on 13 carries on the first two drives, he got just 8 more carries all game!!  All game!!  How is this possible?  So Reich also failed to show up after the first quarter. Though Jax does deserve credit as well.)  Actually, the D came through, but made it entirely too close, partly due to 7 three-and-outs by the O across nine possessions.  That'll sap the spirit of any D, I suppose.

Guts and stomps?  I suspect gutting out a close win over a bad team is predictive of an early January end to the season.

51 Packer defense note

The perceived major weakness entering the season was the lack of talent around Kenny Clark.  And now 10 games in the d line looks pretty good.  Clark has been awesome.  Keke when healthy has continued to provide inside rush and become at least ok against the run.  The Slaton draft pick looks really promising as the 5th round guy is making things happen in his dozen snaps a game.  So much so he looks to have jumped over Lancaster in the rotation.  And then there is Dean Lowry who has been the biggest shocker.  Lowry coming into the season looked to be nothing more than an effort guy.  Every so often you might get a push inside on a pass play.  But mostly he took up blocks.  But something changed end of September or so because Lowry has been periodically during games just wrecking guards.  I stand by my borderline Pro Bowl quality play and in support reference this remark from Justis Mosqueda, writer for Acme Packing Company but truly very knowledgeable about the intricacies of the game.  And no, not just because we agree. People should check out this guys work.  He’s really sharp.  His comment on Dean below 
 

“Kenny Clark says "pass-rush wise" Dean Lowry is playing the best ball in his career. I agree. Been putting up Pro Bowl caliber film the last month”

58 Brief follow up on Lowry

here is a comment from Packer beat writer Tom Silverstein:  Defensive end Dean Lowry looks like someone put a spell on him, he’s playing so well. 

61 53rd/128 DI grade

Starting caliber, hardly pro bowler. The notion he should be cut in the offseason was always silly. The line was already thin and the savings were minimal. He's a fine compliment to Kenny. 

69 Follow up

In reply to by ImNewAroundThe…

I wrote more than once that he was for a stretch playing at a borderline Pro Bowl level.  Did not claim he deserved to be a Pro Bowler

 

FWIW you regularly do this misrepresentation of what a poster writes and then respond to YOUR version.  Just a heads up because it’s poor forum behavior.  
 

If you going to respond with snark or insult just note I have been called just about everything over my long life.  All you will accomplish is that I will know to ignore your posts.  Which if that is your goal just say so

 

Plus it’s complement.  If this was spellchecker that bit you we have all been there 

75 Geez dude.

In reply to by big10freak

First DPF now me. Maybe it's not us.

You can clarify yourself if someone makes a mistake instead of being condescending.  

I know it was a complement I was just pointing out how he's been only alright. I said he was starting caliber, but not pro bowler, sorry. Pretty much everyone has "a stretch playing at a borderline Pro Bowl level"

81 Eye rolling

Strawmen? Gee. Getting worked up and offended over nothing.

First response is to tell them theyre going on the list instead of just continuing with this discussion instead of starting a war.

84 DVOA table Technical Difficulties?

When I hit the click here for the full table I only get to #18 TEN.  I wanted to go back to 2011 and comment on the Tebow comments in the Panthers Posed to Make NFC Playoff Noise article.  When I go to that year I only get the top 18 teams as well.

85 If you point your mouse…

If you point your mouse cursor inside the table, and use the wheel to scroll down, you can see the rest of the teams.

FO is on a never-ending quest to make it impossible to view the entire DVOA table at once, without having to scroll.  Every week, I find a new workaround that allows me to view the entire table at once.  Then, the next week, I find that particular "loophole" has been closed.  It's really starting to get annoying.

87 dDVOA/dt

When I look at the FO DVOA-based SB odds, I wish I could make a bet.  Like all stats, DVOA is backward-looking.  But as we all know, things change.  Why not add a metric that is the derivative of DVOA over time?  I think the prologue to the future is what happened in the past AND how what has been happening has been changing (dDVOA/dt).