AFC Beasts

Miami RB Jeff Wilson
Miami RB Jeff Wilson
Photo: USA Today Sports Images

NFL Week 10 - The Buffalo Bills may have fallen to the Minnesota Vikings in the most exciting game of the year, but we still think they are the best team in the NFL. The Bills are back on top of the Football Outsiders DVOA ratings despite the loss. In fact, their DVOA rating moved up this week.

The top two teams in DVOA both lost this weekend, and they lost despite both teams having a higher DVOA rating than their opponents. The Bills had already moved back into the No. 1 spot before the Eagles played on Monday Night Football, but they opened up a slightly larger lead with the Eagles' loss. Still, the Eagles only drop slightly after losing to the Commanders.

Buffalo finished Sunday's game with 41.7% DVOA compared to -13.0% DVOA for the Vikings. You do not often see a huge gap like that where the team with the better DVOA loses the game. Our Post-Game Win Expectancy formula says that based on the stat splits of this game, we would normally expect the Bills to have won the game 98.6% of the time. That makes this the third-strongest "surprise win" since 2010, ahead of only a Titans-Jaguars game from 2013 and the Miami won over Baltimore back in Week 2 of this season.

Why does DVOA favor the Bills' performance by so much? Well, check out the stat splits:

  • Buffalo outgained the Vikings on a per play basis, 6.75 to 6.18.
  • Buffalo had a much higher success rate on offense, 55% to 41%.
  • Minnesota recovered all four fumbles in this game.

That last one is particularly important because one of those fumbles was the botched center-quarterback exchange for the Bills that resulted in Minnesota's go-ahead touchdown. Those fumbles are almost always recovered by the offense, and I give no credit to the defense on those plays because the defense has nothing to do with a bad exchange. That's all the offense's fault; it's not a forced fumble. (The Vikings are only responsible for the fumble in that they pinned the Bills on the goal line... but we've already given the Vikings' offense credit for that by measuring the success of their previous drive.)

OK, what about the Eagles? That game came out a lot closer, the rare game where both teams had significantly positive DVOA. The Eagles were at 25.0% DVOA and the Commanders at 16.8% DVOA.

  • Philadelphia outgained the Commanders on a per play basis, 5.91 to 4.34.
  • Philadelphia had a slightly higher success rate on offense, 54% to 52%.
  • Washington recovered four of the five fumbles in this game, including all three Philadelphia fumbles.
  • But -- here's the biggest difference between this game and the Bills-Vikings game -- the Commanders ran 81 plays compared to just 47 for the Eagles.

The Commanders had only one drive of three plays. The Eagles had five different drives of just three plays, not counting their final desperation play. (I say "three plays" here because these were not all three-and-out drives; the first Eagles drive after a Washington fumble scored a touchdown on the third play.) Because of that huge discrepancy in the number of plays run, the Post-Game Win Expectancy formula favors the Commanders, suggesting that we would expect Washington to win this game 58% of the time. But it was very close, and that's before we get into things we don't account for in DVOA like completely missed facemask penalties.

The moral of the story here is that the Bills and Eagles are still really good teams. It's still a little strange that DVOA doesn't have the Kansas City Chiefs higher -- they did move up one spot this week to No. 5 -- but a week ago everybody thought that the three best teams in the league were Buffalo, Philadelphia, and Kansas City, and I don't think that this weekend's results should change that at all.

AFC East Looking Great

The Miami Dolphins were one of this week's big movers, climbing from 12th to seventh with a hefty 39-17 win over Cleveland. Shuffling around also moves the New England Patriots up a spot, which puts the entire AFC East into the DVOA top 12 at this point. I thought I would go back and look at the best divisions in DVOA history to see how this year's AFC East compares.

Now, there are two ways to do this. We could look at the average of all four teams in DVOA. However, that would favor divisions that had one superlative team in a year with less parity than 2022. Or, we can look at the average of all four teams' DVOA ranks. That will favor the 2022 AFC East because this year's parity has put all teams a bit closer to average.

Here are the best divisions by DVOA ranks since 2002:

1. 2016 NFC East (6.25)
Cowboys 2, Eagles 6, Giants 8, Redskins 9

2t. 2013 NFC West (7.25)
Seahawks 1, 49ers 6, Cardinals 10, Rams 12

2t. 2022 AFC East through Week 10 (7.25)
Bills 1, Dolphins 7, Jets 9, Patriots 12

4. 2021 NFC West (7.5)
Rams 5, 49ers 6, Seahawks 9, Cardinals 10

5. 2004 AFC East (7.75)
Patriots 2, Bills 3, Jets 4, Dolphins 22

And here are the best divisions by DVOA ranks before 2002:

1. 1983 NFC West (7.25)
49ers 2, Rams 6, Saints 10, Falcons 11

2. 1992 NFC East (8.0)
Cowboys 1, Eagles 2, Redskins 6, Cardinals 14, Giants 17

3. 1991 NFC West (8.75)
49ers 2, Saints 3, Falcons 8, Rams 22

4. 1993 AFC West (8.8)
Chiefs 3, Chargers 5, Broncos 6, Raiders 10, Seahawks 20

What if we looked at the average DVOA instead of the average DVOA rank? Now, the best division since 2002 is the 2004 AFC East, where the average team had 19.8% DVOA. The current average for the AFC East is 16.6%, which would rank fifth since 2002. The best division prior to 2002 by average DVOA rating instead of rank is the 1992 NFC East at 14.3%.

With the seven-team playoff field in each conference, it is now possible for an entire division to make the playoffs in the same year. Right now, the entire AFC East makes the playoffs in 16.2% of our simulations, and at least three of the AFC East teams make it in 82.1% of simulations.

What about the NFC East? That division is not as impressive as the AFC East in DVOA, but the entire division is also at .500 or above. The entire NFC East makes the playoffs in 12.8% of today's simulations, and at least three of the NFC East teams make it in 88.6% of simulations.

The Vikings, Continued

At this point, we'e gotten to a completely ridiculous point with the Minnesota Vikings and our "Worst DVOA at X-X record" tables. Scroll down the page and you'll see the 8-1 Vikings aren't even in the top half of the league in DVOA this year! The Vikings are the first team to ever have an 8-1 record with a negative DVOA. They are over 10 percentage points worse than any other 8-1 team in DVOA history.

Worst 8-1 Teams by DVOA, 1981-2022
Year Team DVOA Rk Final
2022 MIN -3.2% 17 -- -- -- --
2014 ARI 7.1% 13 11-5 -2.5% 20 Lost WC
2012 ATL 8.5% 10 13-3 12.4% 7 Lost CCG
1982 LARD 9.1% 11 8-1 9.1% 11 Lost Div
1982 WAS 13.3% 7 8-1 13.3% 7 Won SB
1988 BUF 13.6% 6 12-4 13.1% 6 Lost CCG
1984 DEN 13.7% 9 13-3 13.1% 8 Lost Div
1983 DAL 14.7% 3 12-4 15.4% 5 Lost WC
2000 OAK 14.8% 9 12-4 21.0% 5 Lost CCG
1985 LARM 16.6% 5 11-5 19.3% 5 Lost CCG

I added information to that table on how each of these teams did at the end of the season, but I honestly don't think you learn anything about the Vikings from that. The Vikings have nothing really in common with the other teams on that table. All the other teams were better overall and the closest team, the 2014 Cardinals, had just lost their starting quarterback Carson Palmer to a season-ending injury when they were 8-1, which is part of why their DVOA drops for the rest of the season while the other teams on the list stay pretty consistent.

Look at some of Minnesota's other stats and that DVOA ranking does make some sense. The Vikings have been outgained by their opponents on a per-play basis, 5.92 to 5.52. They have a 45% success rate on offense and have allowed a 47% success rate on defense. They do have a +2 interception margin (eight thrown, 10 taken away) but their outstanding +8 turnover margin is mostly the product of recovering 67% of fumbles (not including muffed kicks).

Other advanced metrics do have the Vikings higher than DVOA does. Pro Football Reference SRS, based solely on points scored and allowed plus schedule, has the Vikings tied for sixth. ESPN FPI has the Vikings 10th. EPA per play at doesn't list an overall rating but has the Vikings 12th on offense and 13th on defense.

I did a little bit of research into where the differences are between DVOA and these other methodologies. For the most part, it's hard to tell! For example, let's compare Minnesota's rank in DVOA by down to EPA per play:

  • First down: 19 EPA vs. 21 DVOA
  • Second down: 19 EPA vs. 20 DVOA
  • Third and fourth down: 9 EPA vs. 9 DVOA
  • Overall: 12 EPA vs. 17 DVOA

That's weird, right? The ranks are bsaically the same for each down, but when you put it all together, DVOA just has the Vikings lower than EPA per play.

If you look at the same thing for defense, you find a huge difference on first downs. EPA per play on defense puts the Vikings eighth on first down. DVOA has them 21st. But how much of that strong EPA on defense is caused by Josh Allen's aborted snap fumble on his own goal line and the Vikings recovering it for a touchdown? Again, that's a huge change in EPA that we aren't counting in DVOA at all because it says nothing about how good the Vikings defense is.

Another big gap comes on special teams. does not list EPA for special teams but ESPN FPI is split into offense, defense, and special teams. Their special teams methodology puts the Vikings fifth in the league. Our special teams methodology puts them 25th! That's a sizeable difference, and it seems to be entirely related to the value of Minnesota's punt returns. I need to learn more about how ESPN's special teams methodology works, but Jalen Reagor is in the bottom half of the league in yards per return and has fumbled twice (and recovered both of those fumbles).

Anyway, it's certainly possible that DVOA is underrating the Vikings a little bit. But if you use any kind of play-by-play breakdown instead of just total points scored and allowed, this team is not a top-10 team right now. FPI and EPA per play put them about 12th in the league. That still makes them one of the worst 8-1 teams ever, if not the absolute worst 8-1 team ever. Vegas agrees, making Minnesota a home underdog to Dallas this week.

* * * * *

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, Derrik Klassen's All-32 game preview column, and picks against the spread (now including projected totals for over/unders).

*New: It's not a paywall, except on Mondays! As of this week, we are now posting DVOA data on Monday instead of waiting until Tuesday as in the past. If you want to see data from the current season on Monday, including all of Sunday's games, you will need to be an FO+ subscriber. On Tuesday morning when we post the update with Monday Night Football added in, all of the free stats pages become free again. The exception will be snap counts, which will still be available to everyone Monday but now will be updated earlier.

* * * * *

This is the Football Outsiders Top 16 through 10 weeks of 2022, 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). Click here for the full table.

OFFENSE and DEFENSE DVOA are adjusted for opponent and performance indoors and consider all fumbles, kept or lost, as equal value. Opponent adjustments are now at full strength. 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.

DAVE is a formula which combines our preseason forecast with weighted DVOA to get a more accurate projection of how a team will play the rest of the season. DAVE is currently 24% preseason forecast and 76% actual performance for teams with 10 games, and 36% preseason forecast and 64% actual performance for teams with nine games. Cleveland DAVE rating is adjusted to reflect Deshaun Watson returning from his suspension in Week 13. Indianapolis DAVE rating is higher because it is now once again based on a projection with Matt Ryan at quarterback.

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>

1 BUF 35.0% 2 26.7% 1 6-3 14.7% 5 -16.9% 1 3.4% 7
2 PHI 29.7% 1 22.4% 2 8-1 18.7% 4 -13.1% 5 -2.1% 24
3 BAL 28.6% 4 20.5% 3 6-3 19.3% 3 -2.5% 12 6.8% 1
4 DAL 25.8% 3 18.9% 4 6-3 5.5% 12 -15.9% 2 4.3% 3
5 KC 19.3% 6 17.0% 5 7-2 26.1% 1 6.7% 22 -0.1% 18
6 SEA 13.4% 5 10.3% 7 6-4 11.6% 7 1.7% 18 3.5% 5
7 MIA 13.2% 12 9.2% 9 7-3 25.2% 2 6.8% 23 -5.2% 32
8 TB 12.2% 10 10.0% 8 5-5 1.9% 16 -10.3% 7 0.0% 17
9 NYJ 12.2% 8 9.0% 10 6-3 -1.0% 19 -11.2% 6 2.0% 8
10 CIN 11.7% 7 11.1% 6 5-4 7.7% 8 -7.1% 10 -3.0% 28
11 SF 10.3% 9 6.5% 11 5-4 2.9% 14 -9.3% 8 -2.0% 23
12 NE 5.9% 13 4.4% 12 5-4 -13.5% 26 -15.9% 3 3.4% 6
13 TEN 4.0% 17 1.6% 14 6-3 -5.5% 21 -8.0% 9 1.5% 11
14 GB 2.2% 21 3.1% 13 4-6 7.2% 9 0.4% 17 -4.5% 30
15 CLE -1.7% 11 -1.7% 16 3-6 13.3% 6 14.2% 31 -0.9% 20
16 JAX -3.2% 14 -7.1% 24 3-7 5.9% 10 9.8% 27 0.7% 15


102 comments, Last at 18 Nov 2022, 1:28pm

#1 by laflamablanca87 // Nov 15, 2022 - 3:38pm

Genuinely curious as to why the Bills defensive DVOA was -15.0% in the game while giving up 6.18 yards per play.  Is DVOA giving less value to Minnesota's explosive plays?

Points: 0

#2 by IlluminatusUIUC // Nov 15, 2022 - 3:38pm

What does DVOA do with a team like Buffalo, that can string together 9 amazing plays followed by 1 brain-melting error and do it multiple times a game? It's wild now that the perception has flipped and analytics love Allen's offense but the tape is concerning.

Points: 0

#7 by Aaron Schatz // Nov 15, 2022 - 4:22pm

Two picks, two fumbles by Minnesota, although the Bills only get credit for one of those fumbles because one of them was the aborted play (incomplete backwards pass, I think).

Points: 0

#12 by jmaron // Nov 15, 2022 - 4:46pm

so Buffalo defence gets credit for the right guard stepping on the QBs foot which results in the QB fumbling trying to pitch the ball to the RB?

I think Minn has won bc of turnover advantage and they've have started on their own 31.4 vs 26 yd line for opps. They also have 124 additional yards on penalties. They also have 10 more red zone entries compared to their opponents

Points: 0

#63 by Aaron Schatz // Nov 16, 2022 - 9:59am

There are two fumbles. One is listed as an aborted play. The Bills don't get credit for that one, in the first quarter. They do get credit in overtime because that's listed as a sack.

You're right about penalties, that's something in Minnesota's favor that isn't (for the most part) included in DVOA because it is less predictive than yardage on regular offense/defense.

Points: 0

#71 by jmaron // Nov 16, 2022 - 12:28pm

In reply to by Aaron Schatz

5 yards of start position a game seems like a bigger deal to me. Not sure how it's happening but that 50 to 60 yards a game.

Points: 0

#18 by PirateFreedom // Nov 15, 2022 - 6:01pm

dvoa does favor consistent gains over big plays. a longer play is better but the DVOA  increase is less than linear.


edit: sorry reply went wrong place and I can't find delete

Points: 0

#36 by Pat // Nov 15, 2022 - 8:12pm

In reply to by PirateFreedom

DVOA is per-play. A team that scores a long TD every play is the best team ever.

The only reason you might say it "favors" consistent gains is because long gains aren't all that predictive and don't happen often.

Points: 0

#20 by kamso // Nov 15, 2022 - 6:21pm

Wasn’t it 2 picks and 2 fumbles? That’s a lot of potential turnovers

Points: 0

#38 by owleabf // Nov 15, 2022 - 8:25pm

The two fumbles were:

-a pitch to the RB where the RG stepped on cousins foot, nobody within five yards is cook 

-a slightly backwards swing pass that was easily covered by cook with nobody close enough to contest it


The picks were both terrible throws.


Points: 0

#39 by owleabf // Nov 15, 2022 - 8:25pm

The two fumbles were:

-a pitch to the RB where the RG stepped on cousins foot, nobody within five yards of cook 

-a slightly backwards swing pass that was easily covered by cook with nobody close enough to contest it


The picks were both terrible throws.


Points: 0

#21 by Will Allen // Nov 15, 2022 - 6:31pm

I believe that any gain over 40 yards starts to be discounted, because gains of 40+ yards are not predictive. That's no doubt true in general, but I think that's not true with some particular players. An offense with Tyreek Hill, for instance, will actually be a little better than what DVOA indicates, in my opinion.

Points: 0

#79 by ChrisS // Nov 16, 2022 - 1:45pm

Also long gains are field position dependent. The exact same play that gets you an 80 yard TD from the 20 only gets 30 from the other 30.

Points: 0

#82 by siggyllama // Nov 16, 2022 - 3:52pm

Isn't the first play more valuable precisely because it was from your own 20? I guess that's outweighed by the lack of predictive power.

Points: 0

#93 by ChrisS // Nov 16, 2022 - 6:21pm

Yes it is more valuable DVOA takes into account down, distance to go and filed position. But a 80yd TD is less than 4 times more than a 20 yd TD. Though I believe red zone plays get a bonus.

Points: 0

#25 by Kaepernicus // Nov 15, 2022 - 6:50pm

Funky site stuff going down. Elon buy FO too?

Points: 0

#3 by Aaron Brooks G… // Nov 15, 2022 - 3:46pm

But how much of that strong EPA on defense is caused by Josh Allen's aborted snap fumble on his own goal line and the Vikings recovering it for a touchdown? Again, that's a huge change in EPA that we aren't counting in DVOA at all because it says nothing about how good the Vikings defense is.

It's interesting to observe that play wasn't the biggest delta-EPA play in the game -- Cook's TD was. 6.85 vs 6.32 -- by virtue of being at the 1, the Bills had a negative expected EPA at the snap.

Because DVOA discounts fumbles and caps big plays, it discounts both plays. 

Do the Vikings have a big number of big plays? DVOA hates boom/bust teams.

Points: 0

#17 by laurin // Nov 15, 2022 - 5:50pm

I can't remember whether it was DVOA or DYAR that caps plays at 40 yards, but either way: the Vikings have 5 passes of 40+ yards (t-11th) and 2 rushes of 40+ yards (t-9th). 7 total isn't much (t-8th with 8 more teams), so they're above average but not some insane outlier like last year's Seahawks who had 15 plays of 40+ yards on just 954 plays (Vikings have 7 in 590).

Points: 0

#31 by owleabf // Nov 15, 2022 - 7:54pm

Do the Vikings have a big number of big plays?


The Vikings have been doing a lot of failing on early downs, then throwing for 12 yards on 3rd and 8.

Points: 0

#77 by nosoop4u // Nov 16, 2022 - 1:24pm

I'm not seeing what you're seeing. 

Per, Viking are tied for 20th in 20+ yard plays, and tied for 11th in 40+ yard plays.  I can't find anyone who publishes yards to go for 3rd down stats, that would support your hypothesis.

Their variance, as I see it, comes in the fact they seem to be successful early and late in the 1st half, and in the 4th quarter.  That 2nd half of the 1st quarter, the 1st half of the 2nd quarter, and most of the 3rd quarter is just an offensive lull.  They have 6 Q1 TDs, 7 Q2 TDs, just 3 Q3 TDs, and 10 Q4 TDs this year.  Their 1st Q3 TD was in week 7.

Points: 0

#65 by takeleavebelieve // Nov 16, 2022 - 10:24am

What even is a “boom or bust” team? Like the exact specific criteria. A model can’t hate something that is undefinable. 

Points: 0

#67 by nosoop4u // Nov 16, 2022 - 11:25am

Teams with larger standard deviation of play success/value.

For explanation purposes, let's simplify play success to yards per play...

Basically, if 2 teams both average 5 yards/play, but one of them has mostly 4, 5 and 6 yard plays, and the other has mostly 0, 1, 9 and 10 yard plays, the 2nd team would be considered boom or bust.  If you take the standard deviation of all their plays, you can quantify the boom-or-bustiness of the team.

Also, the opposite of "boom-or-bust" would be "consistent," which would be defined by a low standard deviation.

Points: 0

#69 by takeleavebelieve // Nov 16, 2022 - 11:33am

1) Standard Deviation/z-score methodology assumes normality, which I’m not sure is the case here. It may not be appropriate to use. Especially with passing plays, where you will have a huge cluster of 0 yard plays, ie incompletions. 

2) Assuming the data is normal, you still didn’t answer the question. What are the exact ranges that makes a team “boom or bust”?

Points: 0

#70 by KnotMe // Nov 16, 2022 - 11:59am

Well, if your average play is X yards, you could define boom-or-bust as, over 50% of plays are 1 standard deviation from the mean. Not sure how interesting that would be and it's difficult to deal with the cluster at 0. Maybe something like "over X%(50?) pf  plays are either under 2 yards(bust) or over Y yards(boom). Hard to define what the Y should be. 10? 5? 8? more?

Points: 0

#72 by takeleavebelieve // Nov 16, 2022 - 12:34pm

This is exactly the point of the thought exercise - it’s easy to give labels like “boom or bust” without actually having a definition for them. Until you define the Xs and Ys, it’s all just feelings and narrative. 

Points: 0

#73 by RickD // Nov 16, 2022 - 12:47pm

"you could define boom-or-bust as, over 50% of plays are 1 standard deviation from the mean."

I'd also want to say that the variance is interesting.  If the mean gain is 4 yards and the standard deviation is only 1 yard, that doesn't feel like "boom or bust" at all. 

I could keep the original property while scaling the numbers down as much as I wanted.  "Fizzle or bust" comes to mind.

Points: 0

#88 by KnotMe // Nov 16, 2022 - 5:33pm

Maybe something like this:

It's running backs, but you could use the failed completions thresholds for passing plays instead of success rate. Basiclyn high rates of explosive or failed plays. 

Points: 0

#74 by nosoop4u // Nov 16, 2022 - 12:58pm

1) I don't see a problem with that.  Every incompletion is effectively a bust.

2) Don't view it as a binary ("this is or isn't a boom-or-bust team"), but as a scale - just rank each team by their stddev and the ones with the highest are the "most" boom-or-bust and the ones with the lowest are the least boom-or-bust, or most consistent.  If you want, break the range into rankings by quintiles or by values of stddev, to create tiers of boom-or-bust.  Once you have enough statistical data over several years (or decades), then we can match up statistics to what we see and perhaps have a specific number be the threshold..

Points: 0

#78 by takeleavebelieve // Nov 16, 2022 - 1:33pm

Just a friendly FYI, a fundamentally incorrect approach can’t be hand-waived away. It completely invalidates the whole exercise. The onus would be on you to prove the normality of the sample before proceeding to use standard deviation or z-score. 

Points: 0

#84 by nosoop4u // Nov 16, 2022 - 4:52pm

You know what, you've convinced me... I guess I can't quantify what makes a boom or bust team.  But I can quantify what makes an annoying conversation.  And stop talking about my onus!

Points: 0

#96 by takeleavebelieve // Nov 16, 2022 - 10:56pm

As a wise poster recently wrote: 

It's your argument. It's your burden to prove it.

\spoilers: you can't prove it

Points: 0

#97 by Aaron Brooks G… // Nov 17, 2022 - 8:33am

*snaps on gloves*
This is going to hurt you more than it hurts me.

Points: 0

#4 by theslothook // Nov 15, 2022 - 3:54pm

We seem headed for a repeat of last year with Buffalo gagging away the much needed home field advantage. I saw the Bills explicably get destroyed by the Colts a year ago; a game that was even more embarassing than the score would indicate. And I was mercifully spared from watching the Jags Bills game. 

You can feel the tension surrounding Buffalo and the enormous expectations weighing over the organization and the fanbase. I felt it during the 05 season with the Colts, where every week felt like either doomsday or relief. Not everyone gets an 07 style vengeance tour over the league. 

On a more serious note, the Bills have in theory all the ingredients of a modern NFL team. They have horses on defense and perhaps the best pass offense in the league. The run defense is suspect, but if you are going to be weak at something, you'd take run defense every time. So on paper, this is still the team to beat. Maybe I am just projecting this, but they feel more fragile than a team this strong should. 

Points: 0

#9 by IlluminatusUIUC // Nov 15, 2022 - 4:26pm

Maybe I am just projecting this, but they feel more fragile than a team this strong should. 

Buffalo can't reliably close games on defense, nor finish in the red zone on offense. Aside from blowouts to the Titans and Colts, over the last few years almost all our losses have generally come down to turning 7 into 3 or 3 into 0 once inside the 20. Allen has 5 (5!) ints in the red zone since the bye, and two games we lost by 3 each.

Points: 0

#27 by Kaepernicus // Nov 15, 2022 - 7:08pm

They are a very weird team. They are the worst team in the league when it comes to turnovers on offense. They are the 3rd best team in the league when it comes to creating turnovers. I see mean regression happening and barely budging their TO differential because of those insane splits. The crazy thing about their defense is they are in the bottom 4 teams in the NFL when it comes to average time and plays per possession. Basically it looks like their offense is going to get better and the defense is going to get significantly worse if the turn over numbers regress to the mean at the same time. They will still be good because they are +1.3 net yards per play. They are catching a bit of the 2021 Patriots analytical alchemy and may be overrated.

Points: 0

#100 by JacqueShellacque // Nov 17, 2022 - 10:30am

I think all that can be said of them is that they're high variance, with the 'top' being very high, and the 'valley' being about NFL average. This means they could win it all (regardless of whether or not they win the division and get the bye), or they could be one and done (again, regardless of whether they win the division or get the bye). Since last January, it's really been about what happens this coming January, and that may be how the McDermott/Allen years go. 

Points: 0

#64 by takeleavebelieve // Nov 16, 2022 - 10:15am

They remind me a lot of the Manning/Dungy Colts - an outstanding pass offense that gets ahead quickly, and a defense built around defending the pass to protect those leads. That’ll win you 12-14 games and maybe even a SB ring. But occasionally you run into a bad matchup or just come out flat on offense, and the whole thing falls apart. 

The other worrying trend is that Allen isn’t taking care of the football - 10 INTs (leads the league) plus 8 fumbles. I have to wonder if Daboll’s departure has anything to do with that. 

Points: 0

#75 by IlluminatusUIUC // Nov 16, 2022 - 1:14pm

Oddly enough, the Bills have only "come out flat" on offense three times: LA, Baltimore, and KC, in which we had 10 points at the half each time. And all three of those ended in wins. The games we've lost we had 2nd half leads we couldn't protect. The last time we came out flat and stayed flat long enough to get boatraced was week 11 vs. the Colts last season.

Points: 0

#5 by Romodini // Nov 15, 2022 - 3:56pm

      completely missed facemark penalties

When you play sloppy like the Cowboys, you have to expect to get ref'd like them.

The league doesn't care about questionable officiating at all. The Cowboys sent tape to the league for an explanation regarding their pass rushers being put in blatant choke holds and arm bars in the Lions game and got no response. Next week against the Bears, more of the same obvious holding. Here's the Bear's game. At 0:51 you can see Parsons in an arm bar, at 4:32 you can see another rusher in an obvious arm bar, and at 9:25 you can see Parsons yanked forward into the ground. 

And of course you have the Bills and Vikings game, where the booth review declined to review the Gabe Davis "catch". Everyone else can see it in replay, why can't the booth?

The players obviously don't like the terrible officiating, but I guess the players association is incapable of doing anything about it. 

Points: 0

#8 by Aaron Brooks G… // Nov 15, 2022 - 4:22pm

The facemask was a reference to the Eagles game, where the 4th quarter turned on Goedert being facemasked to the ground and "fumbling." As a result, a guy heading to an All-Pro year is going to miss extended time.

I guess after the Super Bowl, torquing a guy to the ground by his face isn't a penalty anymore. 

Points: 0

#11 by Romodini // Nov 15, 2022 - 4:29pm

I'm aware of that. The Eagles played a sloppy game like the Cowboys did, hence the reference to letting bad officiating impact a game.

Points: 0

#29 by Kaepernicus // Nov 15, 2022 - 7:43pm

Welcome to the club buddy. Nick Bosa has been getting held continuously since the second half of his rookie year. Micah is going to have to figure out how to beat holds and double teams just like the rest of the elite pass rushers in the NFL. The issue right now is the gap between the top pass rushers and offensive tackles is so large that refs have literally just stopped calling everything against the best in the league. Bosa got held the entire SB against Chiefs and drew almost no calls. Micah has to learn how to deal with it before he takes the next leap because the NFL is not going to let monster pass rushers destroy fantasy football leagues around the country.

Points: 0

#35 by BigRichie // Nov 15, 2022 - 8:08pm

Offensive linemen were using their hands back in the 60s. (read Instant Replay; John Madden also remarked on that while having a conversation with Vince Lombardi)

Just because the NFL blocking rules aren't the same rules you learned back in 5th grade flag football ... 

Points: 0

#40 by Kaepernicus // Nov 15, 2022 - 8:29pm

The technique is to keep your hands inside and grab the shoulder pads without ripping jersey or stopping momentum when they get around you. When they start beating you the move is to try to push them outside/inside far enough to let the QB/RB get away. What I am talking about is incredibly strong/fast players blowing off a lineman with a swim/rip move and getting stopped when the blocker is beside or behind them. Today you can keep holding because the odds of getting called are much lower. The stuff that happens to Bosa is a direct result of his power and acceleration. He sheds blocks quickly, has incredibly powerful hands, and has twitchy acceleration. This makes him seem barely affected by blatant holding. It's very similar to LeBron/Shaq in the NBA getting less calls than they should because they do not demonstrate the proper reaction to draw the foul even though the impact/position of the offender is the same. Parsons is so fast and powerful that he is going to get held and not get called. It's different rules for different players because they are truly elite. At the moment the only positions that can exploit these athletic mismatches in the NFL are on offense or in the secondary.

Points: 0

#6 by theslothook // Nov 15, 2022 - 4:09pm

I almost put actual money on the under for the Rams. The entire offseason vibes felt like a team treating this season as a victory lap. I ended up not chancing it because the leftover talent is so incredible that I chickened out. That said, while I anticipated a down year, I did not anticipate a trainwreck season. 

Consider that this team's dvoa is worse than the lowly disappointing Raiders. Their defense is average, but the offense and special teams are both hideous. You close your eyes and this team looks an awful lot like the LARM team that dissapointed following the SB loss. Except even that team with its 16th ranked offense is miles ahead of this team, despite Goff's reputation buried in mud by that point. 


Points: 0

#13 by Joey-Harringto… // Nov 15, 2022 - 4:57pm

Not that either team will accomplish much this year, but I find it humorous that Goff has a +17% passing DVOA compared to Stafford's -9.6% (to be fair, Goff has a very good O-line, and Stafford's is trash), and the Rams' pick they traded away to the Lions has a decent shot of being higher than the Lions' own pick.

Points: 0

#14 by theslothook // Nov 15, 2022 - 5:13pm

It's all rendered moot now because they won the Sb, But I thought the decision to trade Matthew Stafford was a fascinating insight into perceived value.

At that point, almost for opposite reasons, there was uncertainty about what Matthew Stafford and Jared Goff were as quarterbacks. Stafford's tenure with the Lions teased of a possible higher upside while Goffs comparatively rosier surroundings warned of a dangerous downside. That meant it took two first rounders plus a downgrade in age to trade lower downside for greater upside.

No one will care now because of the results, but I suspect in reality Goff isn't all that dissimilar from Stafford. I think Stafford is better and by a non-trivial amount, but not enough to justify two first rounders.

I am not sure what everyone's opinion on Goff is, but I feel more confident that he's very much in that volatile tier 3 QB mix as opposed to being in the mitch Trubisky camp of quarterbacks.

Points: 0

#15 by Joey-Harringto… // Nov 15, 2022 - 5:28pm

"It's all rendered moot now because they won the Sb"

Flags fly forever as they say.  You can argue it's still bad process, though.

"I am not sure what everyone's opinion on Goff is, but I feel more confident that he's very much in that volatile tier 3 QB mix "

Yea, I agree with this.  I thought the Bears game was a perfect microcosm of the Goff experience.  Had good protection, and thus had a very good game, but threw the most horrific red zone interception you'll ever see, and was fortunate to have it negated by penalty.  Productive in the right circumstances, but can't really be trusted to be a long-term solution at quarterback.

Points: 0

#19 by greybeard // Nov 15, 2022 - 6:16pm

Yes. But you are comparing him to Stafford who is a prominent expert on throwing pick-sixes.  
In my opinion there is very little that differentiates Goff and Stafford other than Stafford got lucky that Tartt dropped an easily interception and he played against a mediocre Bengals in SB vs Goff had to play against a good Patriots team. 

Points: 0

#26 by Will Allen // Nov 15, 2022 - 7:07pm

Stafford never had a bad healthy season after his rookie year, until this year at age 34(assuming he's healthy now), and racked up several top 10 DVOA seasons with the more frequently than not poorly coached Lions. This is the 1st year Goff has looked like he can be better than his context. If I had to bet meaningful money on which guy to draft at age 22, it's not a close call.

Points: 0

#30 by theslothook // Nov 15, 2022 - 7:51pm

I'd be curious how we would view Jared Goff if they had won that super bowl with him doing as little as he did on offense. 

I think Stafford's reputation is now cemented because of that super bowl win. And that's actually kind of reasonable to an extent because Stafford, for all his faults, is much better than Flacco or Brad Johnson or certainly Trent dilfer. He's always been a good quarterback. Maybe he's functionally the same as Eli Manning, although I think Eli was a better player under pressure ( I don't mean in a clutch sense but literally while under pressure).

I just think we've seen enough of Stafford at this point to cement the fact that he's not really more than a tier 3 quarterback. I think he might be on the higher end and Jared Goff might be on the lower end but they both live inhabit the same neighborhood.

Maybe that's where you and I differ because I do think they're much closer than people otherwise think.

Points: 0

#34 by Will Allen // Nov 15, 2022 - 8:04pm

Stafford is 34. Even in the era of 40 year old qbs, he's nearly certain to not be as good as he was when younger.

Brad Johnson was better than you remember. Flacco was far inferior.

Points: 0

#57 by Noahrk // Nov 16, 2022 - 8:30am

I thought Stafford was disappointing last year, too. But how much is the elbow? Mustn't forget that pesky tennis elbow.

Points: 0

#66 by BJR // Nov 16, 2022 - 11:12am

The off-season chatter suggesting the freshly-ringed Stafford should be a HOF candidate was absolutely stupid/overblown, but there's no doubting he was good last year. 4th in QBR, 6th in DYAR. And he was very good in the playoffs too. 

Points: 0

#68 by Noahrk // Nov 16, 2022 - 11:28am

His numbers aren't bad, but he was in a very good offense (and team) and he made some really bad throws at times. I expected him to be more reliable. Maybe even look like a proper HoF. I was a bit disappointed.

Points: 0

#62 by JonesJon // Nov 16, 2022 - 9:49am

When you adjust each of their seasons for the quality of their OL play Goff and Stafford have incredibly similar production. 

Points: 0

#10 by SportsPhan8 // Nov 15, 2022 - 4:28pm

I love how you have a stat called Dave, reminds me of the cop from Harold and Kumar:

”Kumar? What kind of name is that anyway? What is that, like five Os or two Us?

“Whatever happened to good old American names like Dave or Jim?”

Points: 0

#16 by serutan // Nov 15, 2022 - 5:31pm

In reply to by SportsPhan8

DAVE always bummed me out a bit because there wasn't a companion stat called HAL.

Points: 0

#22 by nat // Nov 15, 2022 - 6:32pm

Now, there are two ways to do this. We could look at the average of all four teams in DVOA….Or, we can look at the average of all four teams' DVOA ranks.

You could do a weighted average, giving less weight to the top team in the division, a little more weight to the second team, etc. A weighting of 1, 2, 3, 4 is easy to explain and even kinda matches one way people think about hard/competitive divisions. When you ask “how much competition did the division winner have?” you look at the other three teams.  When you ask how much competition the second place team had for second place, you look at the next two teams. Etc. 

That would suggest weightings of 0, 1, 2, and 3. But they all are competition for the rest of the league. So adding 1 to each to get weights of 1, 2, 3, and 4 makes some sense. It would judge divisions primarily on having no softies, would reward them for having mostly good or great teams, and give a little bonus for having an elite division winner.

It depends on what you mean by a strong division. There’s a lot more than two ways to think about it.

Points: 0

#48 by rmjohnston // Nov 16, 2022 - 1:24am

Would be to calculate 1+DVOA for each team in the division and multiply the results. Take the fourth root of that result and subtract 1 to rescale to DVOA scale. For any given total DVOA, balance is rewarded and imbalance. For example, Four teams with 0% DVOA result in a weighted average of 0%, while a combination of two team at 50% DVOA and two at -50% DVOA comes to a weighted average of -13.4%. Things get a bit screwy if you have two teams with under -100% DVOA, but that's not a significant worry - extremely unlikely after one game played, ridiculously unlikely after two, unimaginable after more - and can be fixed by multiplying 1+DVOA for both teams, taking the negative of the absolute value of the result, and multiplying by 1+DVOA for the remaining two teams.

Points: 0

#54 by nat // Nov 16, 2022 - 6:55am

Geometric mean of 100% + DVOA (which you described) could work.

It’s like asking “how likely is this division to all win their games in a week when  they each play an average opponent?”. That’s a fine way to think about strong divisions.

The playoff odds stuff on FO converts DVOA advantage to a winning percentage, so that could be used instead.

Points: 0

#53 by HitchikersPie // Nov 16, 2022 - 5:06am

I think the harmonic mean might be more instructive in this case, which suggests that among the teams Aaron mentioned:


1. 2021 NFCW 14.7%

2. 2016 NFCE 11.9%

3. 2013 NFCW 11.5%

4. 2022 AFCE 11.2%

5. 2004 AFCE #NUM! (harmonic means only make sense with all +ve numbers and though the Pats-35.0%, Bills-31.4%, and Jets 27.2% were insane in '04 the Dolphins were... not at -14.4%)



Points: 0

#23 by Will Allen // Nov 15, 2022 - 6:48pm

My eyes only differ with DVOA, with regard to the Vikings, in one small way. I see their offense being a little better than what DVOA says, and see their defense as being a little worse. Now, my opinion of their offense may be unduly affected by Jefferson making incredible catches.On the other hand, I really do think their defense has improved. A month ago, bad qbs looked like 44 year old Tom Brady when they played the Vikings. Now bad qbs look like bad qbs; Heinecke would have ranked at the bottom in last week's quick reads, absent a ref blocking a Viking safety, gifting the Commies a 56 yard td. Anyways, if they want to win a playoff game, their defense still needs to get bettter. Stuff like that occasionally  happens between mid-November and January, so who knows?

Points: 0

#33 by Moridin // Nov 15, 2022 - 7:59pm

I'm certainly not going to complain about the rating & ranking. Most decent teams probably have a couple of games a year with such improbable & late game heroics to squeak out a victory, and the Vikings basically have a 7 game win streak based on it. Replay this season w/ the same teams being fielded, and the Vikings could easily be 3-6 or 5-4 or anything really without much of a DVOA difference.

Points: 0

#42 by vikingflagship // Nov 15, 2022 - 9:25pm

i agree the defence is improving.  A big difference is turnovers - Zimmer defences did not emphasise interceptions, they tried to force incompletions. This year the DBs have been getting key, 4th quarter / endzone interceptions. Will it sustain ? who knows ? Big plays in q4 win tight games. 

The run defence has also improved since the start of season, and Hunter is having bigger impact. we hope these improvements come from learning the scheme, but we cannot know. 

On ST, my eye test is we are better than DVOA. Punting , kickoff, and coverage all look eye test ? I don't watch from behind the sofa.... the ST coach is a dude. 

The negative plays - missed XP, missed 50+ FGs have not hurt yet (they are 8-1, after all) .  

Points: 0

#59 by peterplaysbass // Nov 16, 2022 - 8:40am

The anchor on their special teams ranking is the XP/FG value, and you're right, that hasn't hurt the Vikings yet (other than all of the Pepto Bismol and Alka Seltzer they need to ingest).

Points: 0

#87 by jmaron // Nov 16, 2022 - 5:26pm

21 of the last 26 Vikings games have been 1 score games. The 2 last games of 2021 which didn't mean anything were multi score games. 

So in games Viking's fans have cared about, 21 of 24 have been one score games.

The Vikings are literally going to cause heart attacks to their older fans. 



Points: 0

#91 by nosoop4u // Nov 16, 2022 - 6:04pm

As if the elderly didn't have enough to worry about these past 2½ years...

Points: 0

#28 by AFCNFCBowl // Nov 15, 2022 - 7:30pm

the Vikings have had negative DVOA in 6 of 9 games this season (!!!)

Wins by the Vikings with Negative DVOA:

W5 CHI, -25.6%

W4 NO, -22.9%

W10 BUF, -13.0%

W9 WAS, -11.8%

W8 ARI, -3.4%

DVOA also doesn't know the Vikings faced Bridgewater and not Tua (inflates their rating for the MIA game)

by DVOA the Vikings have actually been a disappointment this year (they were projected to be a very good team in the FO almanac).


Points: 0

#32 by Will Allen // Nov 15, 2022 - 7:58pm

From game 2 through 6, the defense was simply atrocious. Indescribably bad. If they revert to that, they could still end up losing 8 games.

Points: 0

#37 by owleabf // Nov 15, 2022 - 8:16pm

As a Vikings fan, a few thoughts:

I get the theory of expecting fumbles to balance out, but I wish there was a way to add context. The two offensive fumbles by the Vikings never had a realistic chance of going to the bills. In both cases it was to dalvin behind the LOS with nobody within five yards.

Take it as you will, but according to reporting from the Athletic the fumbled snap was actually caused in part by a shift from Harrison Phillips in response to their audible on the line. Regardless, if the bills recover that fumble it's a safety and the Vikings still have a very reasonable chance of winning that game.

Overall, I wonder if there's room for thinking about why this year's 8-1 Vikings team is lower rated than last year's 4-5 Vikings team (if memory serves.) The motto on this year's Vikings is "situational matters" and watching the games it feels like there is some truth to it, I wonder if there is a way to account for win probability more than EPA/play. DVOA would have it as clear that the Vikings "should" go 0-4 on this BUF/DAL/NE/NYJ stretch, but I have trouble seeing them losing even two of them.

Points: 0

#41 by Will Allen // Nov 15, 2022 - 8:36pm

Injury luck, fumble luck, and schedule luck are the most significant areas of superiority the current regime has shown, compared to the old, so far. I say that as somebody who was fine with the old regime getting fired.

Points: 0

#45 by owleabf // Nov 15, 2022 - 10:16pm

Injury luck seems to at least have some strategy/skill element to it. The Rams, KOC's previous team, by Adjusted Games Lost were 4th in 2018, 10th in 2019,  2nd in 2020 and 5th in 2021.

In my opinion last year's team folds when they're down 10-27, there's something different this year. We'll see how things play out over the next few weeks, as I mentioned, the Vikes play DVOA's  #4th, #9th and #12th ranked teams. By DVOA's account Dallas "should" blow the Vikings out, let see what actually happens.

Points: 0

#46 by owleabf // Nov 15, 2022 - 10:16pm

Injury luck seems to at least have some strategy/skill element to it. The Rams, KOC's previous team, by Adjusted Games Lost were 4th in 2018, 10th in 2019,  2nd in 2020 and 5th in 2021.

In my opinion last year's team folds when they're down 10-27, there's something different this year. We'll see how things play out over the next few weeks, as I mentioned, the Vikes play DVOA's  #4th, #9th and #12th ranked teams. By DVOA's account Dallas "should" blow the Vikings out, let see what actually happens.

Points: 0

#51 by vikingflagship // Nov 16, 2022 - 2:37am

Is there a way to see 4th quarter DVOA ? Or DVOA by quarter in general.  
It seems like Vikings have played their best football in the 4 quarter. 

Points: 0

#81 by Vincent Verhei // Nov 16, 2022 - 2:46pm

Absolutely! With a subscription to FO+, you can see DVOA splits by quarter, field zone, down-and-distance, and more. More info available here!

Points: 0

#92 by vikingflagship // Nov 16, 2022 - 6:17pm

Found it.  Thanks.  

Points: 0

#60 by peterplaysbass // Nov 16, 2022 - 8:44am

Another area of superiority seems to be whatever is allowing Cousins to take some more risks. They're still calculated (no risky throws on 2nd and 1, for example), but he's not hoping for YAC and settling for a punt by throwing short on 3rd and long as often as in years past.

It's like someone sprinkled in a little 2017 Case Keenum into the mix, but with less arch on the long throws.

Points: 0

#56 by fyo // Nov 16, 2022 - 8:00am

I seem to recall that DVOA doesn't treat all fumbles the same, but perhaps that's just me misremembering.

For fumbles on QB-center exchanges, the probability of the offense recovering the fumble is nearly 75% (2000-2011, On a sack it's pretty much 50/50. Fumbles after a reception are another matter completely, with the defense recovering 60% of the time (same source). I haven't seen a study that looks at backward passes, but fumbles after non-QB runs and forward passes have the roughly same probabilities, so I would use that number (40/60) as a guestimate.

As for providing more specific context to each individual fumble, that's a fool's errand, IMHO. If you provide infinite context, you can predict with near 100% certainty who will recover the fumble, completely removing "fumble luck". I would argue that it is more interesting to optimize for future fumbles, not the specifics of any single fumble. I.e. if a quarterback consistently fumbles in when he gets sacked (hello Tannehill), the average probability of a sack fumble resulting in a turnover is more interesting than whether the specific circumstances of a single sack made it extremely likely that the offense would recover.

Points: 0

#61 by Noahrk // Nov 16, 2022 - 8:54am

I don't understand the idea of wanting to cut the Vikings a break because their fumbles were safe. Did they fumble deliberately, then? Of course not. Fumbles are chaotic plays and there is no plan and no preparation involved. If you recover, that's the very definition on luck, situational probabilities notwithstanding.

Points: 0

#98 by Mike B. In Va // Nov 17, 2022 - 9:45am

Right. Just because these two weren't damaging doesn't predict that the next two won't be.

Points: 0

#43 by big10freak // Nov 15, 2022 - 9:47pm

Sigh.   Back near the bottom 

Points: 0

#44 by theslothook // Nov 15, 2022 - 10:09pm

I am kind of glad that it has because it dispenses this view that special teams is all about coaching and if you just had the right coaching in place, special teams would be fixed. Maybe that's kind of true if a coach devotes more of emphasis to it at the expense of other areas, but not in the way most fans think.

The first lesson you learn as a fan is you aren't the smartest guy in the room. The second lesson you learn is you don't know what you think you know. Unfortunately, I expected FO fans to have passed the first two rules of fight club and some have not. (not aimed at you). 

Points: 0

#50 by Spanosian Magn… // Nov 16, 2022 - 2:18am

I agree with all of this. But for the Packers' situation, specifically, a couple caveats:

1) I got the impression there was a (self-imposed) mandate to use Amari Rodgers as the returner in order to get immediate value (theoretically) out of a project of a 3rd round pick. Unfortunately, he was pretty bad at it - so their return game might improve now that he doesn't have the job.

2) I did a cursory review of Bisaccia's history awhile back, and his special teams units weren't terribly impressive, by and large. That may or may not be his fault - they could just be bad players, as you're getting at - but it's also entirely possible that they simply hired a different coach, not a better one.

Points: 0

#55 by BlueStarDude // Nov 16, 2022 - 7:04am

Bisaccia did fairly well in Dallas. I mean, not as good as John Fassel has been doing. But Bisaccia’s units performed better overall than his predecessor DeCamillis's did. And the special teams tanked when he left and O'Quinn took over.

Points: 0

#47 by GwillyGecko // Nov 16, 2022 - 12:10am

It's so disappointing the way Buffalo has likely blown HFA two years in a row now after beating KC in KC both years. And even beating the Titans this year too.  What a humbug Josh Allen has been the last two games.  

Points: 0

#58 by Mike B. In Va // Nov 16, 2022 - 8:37am

I can't say I'm thrilled, but they put so much of the offense in his hands that you almost have to expect this.

There's a part of me that almost wanted him to miss Sunday's game so we could see if the offensive staff can plan a winning strategy or if they're just counting on the Josh Allen "and a miracle occurs" offense now.

Points: 0

#76 by wludford // Nov 16, 2022 - 1:18pm

You say that the Vikings' defense gets zero credit for the Bills fumble in the endzone.  How much negative credit does the Bills offense get?  Obviously not enough if their overall DVOA is +41.7% for the game.  Fumbling in your own endzone, up 3 points with less than a minute left to play, and failing to recover it, thus allowing your opponent a go-ahead touchdown, should have an off-the-charts negative effect by any measure of efficiency and game situation.

The Vikings defense should at least get some credit for recovering the fumble as well, even if they didn't cause it.

This game tests the limits and baseline values for a metric like DVOA.  You may be tempted to disregard it as an outlier, but a better approach would be to use it to assign better values to extraordinary plays that better align with their impact on the game.

It would be interesting to see, by comparison, what the Bills stop of the Vikings the previous play on 4th down was worth in DVOA.  Because the subsequent fumble essentially reversed that value- at least on the scoreboard.



Points: 0

#80 by Bryan Knowles // Nov 16, 2022 - 1:46pm

The fumble was worth -501.8% DVOA for the Bills' offense.  It was a really bad play!

Points: 0

#83 by CVBoot // Nov 16, 2022 - 4:51pm

I'm pretty sure that's the largest DVOA-related number I've ever seen.

Points: 0

#86 by Lost Ti-Cats Fan // Nov 16, 2022 - 5:19pm

"This game tests the limits and baseline values for a metric like DVOA.  You may be tempted to disregard it as an outlier, but a better approach would be to use it to assign better values to extraordinary plays that better align with their impact on the game."

As I understand it, every year FO evaluates their metrics and attempts to improve their predictability.  If "better values" for these sorts of plays improve the predictability, they'll adjust.

The main problem, though, is that these extraordinary plays rarely repeat.  Sure, giving the Vikings credit for end zone touchdown recovery would improve the explanatory power of this particular game.  But what does it say about future games?  That the Vikings are more likely to win because of their ability to recover non-forced opponent fumbles in the end zone?  The highest correlated factor to explaining the outcome of any particular game is points scored:  it has a 100% backwards-looking explanatory power.  The point of DVOA and similar evaluation methods is to look beyond all of the weirdness that results in points actually scored and identify what part of it translates into future scoring power (or stopping power in the case of defenses). 

Points: 0

#94 by takeleavebelieve // Nov 16, 2022 - 10:49pm

Spot on. According to former FO columnist Bill Barnwell, the WPA swing on that fumble was about 95%, which is the largest in the history of that metric. Skewing models to predict once-in-a-decade events inherently results in them being less accurate at predicting everyday events. 

Points: 0

#85 by Jblades13 // Nov 16, 2022 - 5:17pm

The justification for having the Bills' single game DVOA so much higher than the Vikings for this game only makes sense if you don't think about it at all. 

The difference of 6.8 to 6.1 yards per play seems huge, but ultimately it comes out to the Bills outgaining the Vikings by 5 yards for the game. That would suggest the two teams were much more even than the DVOA output suggests.

For 3 of the 4 fumbles there were no Bills players with a realistic chance to recover. The only one that was close was the Allen fumble in the end zone, it doesn't make sense to treat all of those recoveries as fluke 50-50 plays when only one of them was. 

It may just be worded wrong, but giving the Vikings no credit for that fumble also makes no sense, even if it was a clean snap, (which just for the sake of argument we'll say the Vikings defensive shifts and penetration had no effect on the fumbled snap), the Vikings D line clearly got a lot of penetration, and the play would've likely been a safety anyway which would've been a good play for the Vikings and should warrant at least some credit. 

I'm interested to know how much of a defensive adjustment there is here too, this is a Bills defense that coming into this game (with a small sample size) was allowing 9 points per game and just over 300 yards per game at home. Disregarding the kneel down, the Vikings offense put up 16 points and nearly 300 yards of offense in their final 4 drives. They had 202 yards and 10 points for the rest of the game which is a borderline average outing for teams playing in Buffalo so far this season. 

Even acknowledging the Bills higher success rate, when you look at EPA, that didn't seem to matter, even disregarding turnovers, the Bills Offensive EPA per play was .219 compared to the Vikings .210, if you include turnovers, the Vikings averaged .06 EPA per play compared to the Bills -.07, even if you cap turnovers to an extent, the Vikings would've had a likely had a higher EPA per play.  

It's curious how traditional stats like total yardage, and time of possession show an even game, some more advanced stats like EPA per play show either an even game or a game where the Vikings played somewhat better, all of which coincide with what actually happened in the game, but DVOA says the Bills were significantly better than the Vikings. 


Points: 0

#89 by Lost Ti-Cats Fan // Nov 16, 2022 - 5:40pm

"but DVOA says the Bills were significantly better than the Vikings. "

More precisely, it says the Bills were significantly better than the Vikings at the sort of things that correlate to future success.  Or, to put it another way, DVOA thinks the things that the Vikings did well and resulted in them winning this game were mostly "noise" and non-predictive of future success.

Points: 0

#95 by LionInAZ // Nov 16, 2022 - 10:52pm

In other words, see comment 80 about the monstrous negative value of the Bills fumble in their own end zone. Fumble luck is fickle.

Points: 0

#101 by Jblades13 // Nov 17, 2022 - 7:15pm

The hand-waiving about fumble luck is one of my biggest points. 

The Vikings recovering those first 3 fumbles wasn't luck because there were legitimately no Bills players in the area to recover them. 

Saying botched snaps are recovered 99% of the time by the offense may be true, idk if there's actually data to back that up, but most of those scenarios where there's a botched snap are not identical to the scenario in this game, where the defense is lined up and shooting directly at the center. Treating these fumble recoveries as the Vikings recovering 3 50-50 balls, and 1 1% chance recovery is not good process. 

Points: 0

#102 by Mike B. In Va // Nov 18, 2022 - 1:28pm

You're missing the point. The next three fumbles might have members of the opposing team near them, and would be no more predictive than these three.

Points: 0

Save 10%
& Support Aaron
Support Football Outsiders' independent media and Aaron Schatz. Use promo code SCHATZ to save 10% on any FO+ membership and give half the cost of your membership to tip Aaron.