Week 11 DVOA Ratings
by Aaron Schatz
The prodigal sons have returned! DVOA's favorite team, the Philadelphia Eagles, have finally returned to the top of our ratings this week thanks to a big win over the rival Dallas Cowboys. The Eagles have had the best record in the league for a couple weeks now, and most observers ranked them as the top team in the NFL. DVOA had them just below the top until now. The Eagles were No. 3 last week but slip ahead of New Orleans, which remains at No. 2. Last week's top team, the Los Angeles Rams, drops to fourth after losing to the Minnesota Vikings, who are fifth.
The Eagles are also our new Super Bowl favorites according to the playoff odds report, although their odds of making the playoffs aren't quite at 100 percent yet. The Eagles still miss the playoffs in a grand total of 24 out of 50,000 simulations.
The Eagles are on top of a very packed group of extremely well-balanced teams. Earlier this year, I wrote about a lack of historically great teams, and that 2017 was a particularly condensed year for the ratings. Not anymore. After big wins by the Eagles and Steelers, we now have four teams with total DVOA over 30%. That is very rare. Only twice in the past 25 years have there been at least four teams at 30% or higher as of Week 11. Only five times before this year were there even three teams ranked that high:
- 1996: Green Bay (8-2, 41.1%) / Dallas (6-4, 31.6%) / Denver (9-1, 30.1%)
- 2004: Pittsburgh (9-1, 41.2%) / New England (9-1, 38.3%) / Philadelphia (9-1, 32.2%) / Indianapolis (7-3, 31.1%)
- 2005: Indianapolis (10-0, 33.7%) / San Diego (6-4, 32.7%) / Denver (8-2, 32.4%)
- 2009: New England (7-3, 39.4%) / Indianapolis (10-0, 34.7%) / Philadelphia (6-4, 30.7%)
- 2012: San Francisco (7-2-1, 41.7%) / New England (7-3, 39.3%) / Denver (7-3, 38.4%) / Seattle (6-4, 34.4%) / Green Bay (7-3, 31.1%)
- 2017: Philadelphia (9-1, 34.5%) / New Orleans (8-2, 32.5%) / Pittsburgh (8-2, 31.4%) / Los Angeles Rams (7-3, 31.1%)
- Honorable mention: In both 1990 and 1993, with only 28 teams in the league, the No. 3 team after Week 11 was at 29.9%.
The other notable fact about the top five teams this year, including those four plus the Minnesota Vikings, is how well-balanced they are between offense and defense. All five teams currently rank in the top eight for both units. There has never been a season that finished with five different teams ranked in the top eight for both offense and defense. Obviously, this season might not end that way either. Four different seasons ended with four teams ranked in the top eight for both units, and the most recent was only two years ago. Arizona, Carolina, Kansas City, and Seattle all ranked in the top eight on both sides of the ball in 2015. However, not a single team was this well-balanced last year.
Behind the five well-balanced teams are the significantly unbalanced teams. The New England Patriots move up to No. 6 this week. The Patriots have the best offense in the league, and are now third on special teams. Yet they are still near the bottom of the league in defensive DVOA, even after giving up just 12.5 points per game for the last six games. New England drops a place back to 31st this week, although there's just a rounding error difference between the Patriots and No. 30 Miami. What's going on here? I've devoted a lot of writing to the Patriots defense this year, but let's do it again, since DVOA's rank for the Patriots is so much different from conventional wisdom.
- There's somewhat of a gap between the three worst defenses and the rest of the league, and the Patriots were really, really horrendous over the first four games. So each week, the Patriots' defensive DVOA gets a little bit better, but their rank doesn't improve, so it doesn't look like things are getting better if you only look at DVOA rank. Based on current opponent adjustments, New England had 27.9% defensive DVOA in Weeks 1-4, but that has improved to 11.9% in Weeks 5-11.
- A lot of what looks like defense is actually the offense and special teams putting the defense in excellent field position. The Patriots are the worst defense in the league in terms of yards per drive, but they also have the second-best average starting field position on defense.
- The Patriots have had some really good luck, especially with opponent field goal kickers missing field goals. That means fewer points scored against them. There have also been some well-timed fumbles, such as the Austin Seferian-Jenkins play.
- The Patriots once again have a "bend but don't break" defense. There's almost no year-to-year consistency for defenses that are much better in the red zone compared to overall, but the Patriots seem to be the exception. That's worth its own separate article. The difference isn't as big as you might think this year, though: the Patriots are 17th in DVOA in the red zone.
- We're punishing the Patriots for giving up yardage once the game is effectively over. There's something to this argument. The Patriots currently have 10.6% defensive DVOA in quarters 1-3, but 36.6% defensive DVOA in the fourth quarter. Of course, both of those figures rank 30th in the league. As we often point out: DVOA compares teams to a league-average baseline, so the Patriots in a "prevent defense" late-game situation are getting compared to other teams in a "prevent defense" late-game situation. And our past research has showed that removing or lowering the importance of fourth-quarter plays actually makes DVOA less predictive unless the game is significantly out of hand, by over three touchdowns. But we can always take time in the offseason to research this again.
No. 7 is now Jacksonville, with Baltimore at No. 8. In Weighted DVOA, giving less strength to the first couple games of the season, the teams are switched. Baltimore has now passed the Jaguars as the No. 1 team in defensive DVOA, and the Ravens are also No. 1 in special teams. That's how Baltimore is so high in DVOA even though the offense is still horrid. The Ravens have a more balanced defense than the Jaguars, although Jacksonville has significantly improved against the run since picking up Marcel Dareus. They've gone from 32nd to 24th against the run, and are still No. 1 against the pass. As of right now, the Jaguars would rank as the No. 5 best pass defense ever measured by DVOA, trailing the 2002 Bucs, 1991 Eagles, 1986 Bears, and 1988 Vikings.
Baltimore is No. 2 against the pass and No. 10 against the run. However, the Baltimore defense is also No. 1 in variance. Baltimore has three shutouts this year, but were average on defense in a lot of their other games.
This is where we get into one of the other known issues with DVOA, which also plays a role in Philadelphia climbing to the top spot this week. When we compute opponent adjustments, those adjustments are based on performance over the full season. That creates a complication when major injuries significantly change the quality of a unit. There's a big difference between the Green Bay Packers with Aaron Rodgers at quarterback, and the Packers with Brett Hundley. The adjustments for Baltimore in last week's game are based on a combination of the two. There's a big difference between the Dallas offense with Ezekiel Elliott and Tyron Smith, and the offense without Elliott and Smith. The adjustments for the Eagles shutting down the Dallas offense are based on the full season, and most of those games had Elliott and Smith on the field.
Unfortunately, trying to do different opponent adjustments based on injuries like this presents a number of issues. We've written about this in the past, and I will again now. There are reasons why we don't try to adjust for injuries, and instead tell readers that you simply have to use common sense to understand when the ratings might be heavily influenced by a game against backups.
- Do we only adjust for backup quarterbacks, or for all positions? If we adjust for all positions, what do we do about positions where it is extremely difficult to determine the value of a specific player? We do occasionally make adjustments in the ratings we use for the playoff odds, but this is mostly guesswork except for the quarterbacks.
- If we adjust for more than one player, what do we do about the crossover of multiple players being injured? How do we adjust defenses facing the Dallas Cowboys offense when Tyron Smith comes back but Ezekiel Elliott is still suspended?
- How do we adjust if we only have one or two games of sample size to see how well a team plays without a certain player? What if the backup has a surprisingly good game, like in Week 3 of 2015 when Luke McCown started in place of Drew Brees and completed 31 of 38 passes for 310 yards?
- Even if we could figure out how to answer these questions, we run into the problem of programming the whole thing. We're not really computer programmers around here and we still do a lot of our numbers in very old-school Excel.
So, with that in mind, there's no question that the Ravens are ranked a little higher because they got to shut out two backup quarterbacks. On the other hand, the effect of a single game with the Packers is smaller than you might think, and we probably could have a good debate about whether we should even consider Miami with Matt Moore at quarterback to be an easier offense to play than Miami with Jay Cutler at quarterback. This is going to continue to be a problem because the Ravens will get the Houston Texans with Tom Savage this week, and the Texans opponent adjustments are based in part on games with Deshaun Watson at quarterback. We just need to keep this in the back of our minds rather than taking the numbers as gospel.
No matter whether DVOA is overrating their defense or not, the Ravens are in the driver's seat for an AFC playoff spot. Their remaining schedule ranks 24th in average DVOA of opponent, and four of those six games are at home. The Ravens are very likely to finish with a winning record. Only five AFC teams have winning records right now. The only second-place team with a winning record is Tennessee, and the Titans have a head-to-head win over Baltimore, but the Ravens have a big advantage over anyone else in the wild-card race. The other 5-5 team is Buffalo, which has completely imploded in the past three weeks. After that, we're talking about 4-6 teams, and the Ravens have head-to-head wins over Cincinnati, Oakland, and Miami. So the FO playoff odds simulation gives Baltimore a 74.8 percent chance of making the playoffs, even though there's only a 3.0 percent chance of the Ravens winning the AFC North. If we were to compute Baltimore's defensive DVOA using only Green Bay games with Brett Hundley at quarterback, it wouldn't even drop their playoff odds below 70 percent.
* * * * *
Once again this season, we have teamed up with EA Sports to bring Football Outsiders-branded player content to Madden 18. This year, our content for Madden Ultimate Team on consoles comes monthly, while our content for Madden Mobile comes weekly. Come back to each Tuesday's DVOA commentary article for a list of players who stood out during the previous weekend's games. Those players will get special Madden Mobile items branded as "Powerline, powered by Football Outsiders," beginning at 11am Eastern on Friday. Our stars for Week 11 are:
- G Joe Berger, MIN (HERO): Berger anchors a line that had its 6th game without giving up a sack in Week 11 against the Rams.
- P Chris Jones, DAL: No punts returned, three punts downed inside the 15.
- OLB Matt Judon, BAL: 2 sacks, TFL with forced fumble, tackle to prevent conversion after third-and-4 reception.
- WR Kenny Stills, MIA: No. 2 among WR in DYAR for Week 11 (7-of-8, 180 yards, TD).
- G Xavier Su'a-Filo, HOU: Texans RB had 22 carries, 84 yards and a 68 percent success rate on runs up the middle.
* * * * *
All stats pages should now be updated through Week 11, including snap counts, playoff odds, and the FO Premium DVOA database.
* * * * *
[ad placeholder 3]
These are the Football Outsiders team efficiency ratings through 11 weeks of 2017, 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 to 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.
WEIGHTED DVOA represents an attempt to figure out how a team is playing right now, as opposed to over the season as a whole, by making recent games more important than earlier games.
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>
- NON-ADJUSTED TOTAL DVOA does not include the adjustments for opponent strength or the adjustments for weather and altitude in special teams, and only penalizes offenses for lost fumbles rather than all fumbles.
- ESTIMATED WINS uses a statistic known as "Forest Index" that emphasizes consistency as well as DVOA in the most important specific situations: red zone defense, first quarter offense, and performance in the second half when the score is close. It then projects a number of wins adjusted to a league-average schedule and a league-average rate of recovering fumbles. Teams that have had their bye week are projected as if they had played one game per week.
- PAST SCHEDULE lists average DVOA of opponents played this season, ranked from hardest schedule (#1, most positive) to easiest schedule (#32, most negative). It is not adjusted for which games are home or road.
- FUTURE SCHEDULE lists average DVOA of opponents still left to play this season, ranked from hardest schedule (#1, most positive) to easiest schedule (#32, most negative). It is not adjusted for which games are home or road.
- VARIANCE measures the statistical variance of the team's weekly DVOA performance. Teams are ranked from most consistent (#1, lowest variance) to least consistent (#32, highest variance).
79 comments, Last at 07 Dec 2017, 9:55am
#1 by Tundrapaddy // Nov 21, 2017 - 7:56pm
"C Joe Berger, MIN (HERO): Berger anchors a line that had its 6th game without giving up a sack in Week 11 against the Redskins."
I like to toot the Vikings' horn as much as the next guy, but we should probably point out that you're unlikely to give up sacks against teams that you're not actually playing that week.
#3 by Raiderfan // Nov 21, 2017 - 8:15pm
Maybe the easy and objective way to account for injuries is just to increase the WEI proportion more greatly in favor of more recent games, which would also enable you to more accurately reflect things such as the Elliot suspension. I have no idea what formula you are using now, but given the whole universe is small sample sizes (it is only a16 game season), I would not think that should be a major factor.
#4 by ClavisRa // Nov 21, 2017 - 8:28pm
DVOA is good at one thing, and, as far as I can tell, one thing only: elevating to your attention qualities of teams that other stats (depending on what those stats are) might not illuminate. And that's useful, for sure, because when faced with potential information overload, or aggregate stats, it helps direct you where some good areas for a deeper look might be. But, the stats don't on their own mean anything useful. Telling me that a team like the Pats has a terrible defensive DVOA gives me absolutely no idea of how well or poorly they will perform in their next game, or rest of season, or even how they compare to other defenses. If you're looking at DVOA and saying one team is better or worse than another in defense or offense because of their rankings, you're making a huge mistake.
#68 by Hoodie_Sleeves // Nov 22, 2017 - 9:58pm
"much better than W-L" is a ridiculously low bar.
And I'm not sure, honestly, that defensive DVOA actually does clear that bar. FO has long maintained that defenses are less consistent year to year than offenses - and I'm not sure that's true. I'm starting to think that DVOA just does a poorer job measuring defenses, and there's way more noise.
#71 by sbond101 // Nov 23, 2017 - 8:43am
The right bar for DVOA overall is to compare it to the money-weighted consensus of the crowd (the midpoint betting spread - the win/loss odds rather than the point spread as everyone is familiar with the infamous back-door-cover). I'm not sure how that compare comes off in large part because historical betting data isn't readily available (and I have a day job ;)). It's also true that game circumstances should be a layer fitted on top of DVOA, so you'd have to add that layer on top before doing the compare. From general observation it's not clear whether DVOA better fits W/L data then the money weighted consensus from the pre-analytics era (i.e. before FO could influence that consensus) - it seems close. What that means is that the eye-test from an educated observer should be taken pretty seriously when it contradicts what DVOA indicates, but .
Evaluating the predicative power of off/def/spec DVOA is a lot more difficult. Unlike with overall DVOA there is no ultimate yardstick for what qualifies as "good" or "bad" work in a phase of the game. Win probability contribution has some promise as a yardstick, but anyone whose looked at PFR after watching a game knows how BS that matric can be in non-typical circumstances (i.e. BB "taking the wind" in overtime). I get where your comment on defensive DVOA is coming from, I just don't know how to even go about evaluating it objectively.
#33 by Aaron Schatz // Nov 21, 2017 - 11:27pm
No. I screwed it up when I produced the table, of course, like I always seem to do. I will fix the table now. I'm sorry we still do things with so much manual labor instead of having everything run automatically on servers, and so there are always human errors.
#7 by RickD // Nov 21, 2017 - 8:51pm
I have to admit, I was looking forward to the rationalization for assigning the Patriots a poor defensive DVOA even though they held a top 10 offense to 8 points. And you didn't let me down.
You say that the defense is just taking advantage of the achievements of the offense and the special teams when it allows long yardage drives without giving up points. That's like saying the American military is just taking advantage of the oceans when they kept the Axis powers at bay. The point of defense isn't to minimize yardage, it's to minimize points allowed. And when a team has a large lead, like the Patriots did most of the day on Sunday, allowing long, slow drives really doesn't hurt the team much.
It seems like DVOA just isn't doing a good enough job accounting for game situation - either placement on the field or the lead at the time.
Basically you have a model of how teams should play defense - a defense that is wired pretty much the same under all conditions. But the Belichick defense is very much attuned to the scoreboard. When they have a large lead, the defense changes to focus far more on preventing big gains than anything else. When the Raiders are down 30 points in the fourth quarter and take over five minutes to get one TD, Belichick doesn't care at all. It's like watching a pawn moving toward the back rank to be promoted when a checkmate is imminent on the other side of the board. It just doesn't matter.
I'm sure I'm oversimplifying what's wrong here, but something is definitely wrong. And I'm not talking about the early season games when the Patriots were giving up 30 points per game. I'm talking about how you manage to downrate them now in spite of their dominant performance over a Raiders' offense that still rates highly by your metrics.
#13 by Mountain Time … // Nov 21, 2017 - 9:11pm
The numbers downrate them, as I'm sure you know, not the writers. If you don't know that, please re-write your comment to fit the complaints template above xD
I think it's a case of the model not being optimized for this one outlier, which isn't something to criticize.
#36 by RickD // Nov 22, 2017 - 12:03am
"The numbers downrate them, as I'm sure you know, not the writers. "
The writer is being criticized for his rationalization. Hence the word "rationalization".
"I think it's a case of the model not being optimized for this one outlier, which isn't something to criticize."
I would disagree. Criticism is exactly how models are improved.
This is not merely an "outlier" it's an absurdity. Since we do not know the exact details of how DVOA works, we are reduced to judging it by the sausage it produces.
Ordinarily, if a flaw is discovered in a model, the model is deemed a fit subject for modification. After all, what is the purpose of modeling if not to best fit the reality of a situation?
When we have a team that is holding its opponents to low scores consistently, as the Patriots did last season and as they've been doing for the past six games, and we are told that the defense is below average and worthy, the model is diverging from reality. A more appropriate response not based on attacking the messenger would be to try to figure out exactly what is going wrong.
#60 by Aaron Schatz // Nov 22, 2017 - 1:56pm
The model is always a fit subject for modification. The problem is twofold.
1) Time limitations. It takes a lot of time to try to test and tweak, and we have less and less time to do so with more and more other responsibilities around here.
2) Results. Sometimes, we have a hypothesis, we test that hypothesis, and we find out a hypothesis based on a single team is wrong when we try to apply it to all 32 teams. It doesn't mean we won't try again, perhaps try to apply adjustments in a different way. But the goal of DVOA is to get the most accurate ratings for the league as a whole, not for only the New England Patriots.
Criticism always seems to suggest that we think everything we do around here is perfect. But in fact, I hope it is obvious that things are the opposite of that: in some ways, I use these weekly DVOA commentaries as an ongoing discussion about the flaws in the methodology, because I also get confused when we get unexpected results and I also want things to be as accurate as possible. It's hard to have that open ongoing discussion in the face of criticism that is angry and snarky instead of positive and constructive.
By the way, on the subject of the Patriots defense by quarter, I split that by Weeks 1-4 compared to Weeks 5-11.
If we only look at Weeks 5-11, the difference is smaller. The Patriots' defensive DVOA in quarters 1-3 is 6.7%, 24th in the NFL. In the fourth quarter it is 24.0%, 25th in the NFL.
In Weeks 1-4, that split is 16.3% for quarters 1-3 (29th) and 56.6% for the fourth quarter (30th).
The Patriots' extremely poor defensive DVOA in the fourth quarter is heavily powered by the first game of the season when the Kansas City Chiefs scored 21 points in the final 15 minutes.
For Weeks 5-11, the worst fourth quarter for DVOA came against Tampa Bay in Week 5. In that game, the Buccaneers had a 74-yard touchdown drive in the final five minutes to make the score 16-14 Patriots, and then had an additional 56-yard drive in the final 1:10 that ran out of time 19 yards away from a game-winning score. It's hard to argue that the Patriots' defense in those final five minutes can be easily written off as "prevent defense." But that was also six weeks ago, which fits the narrative that the Patriots defense has improved over time.
In Week 11 against Oakland, the Patriots' defensive DVOA in the fourth quarter was 9.9%, which, given the small sample size of a single quarter, is very close to average.
#69 by Hoodie_Sleeves // Nov 22, 2017 - 10:02pm
" But the goal of DVOA is to get the most accurate ratings for the league as a whole, not for only the New England Patriots."
The problem is that you're doing this by overfitting everything, so you end up with a whole bunch of nonsense, and anything that pulls the curve towards the middle is "more predictive"
#16 by Will Allen // Nov 21, 2017 - 9:19pm
These comments always sound as if Aaron isn't open to improving the model, and I'm pretty certain that isn't the case. Actually doing it, I suspect, for all teams, over thousands of plays, is rather problematic. More than a third of the season remains. Usually, when I've noticed something that DVOA is missing, by seasons end what I think it is missing, if it continues, tends to get incorporated.
#17 by Hextall_27 // Nov 21, 2017 - 9:30pm
Tom Brady is the worst DVOA rusher in the last 2 minutes in NFL History!
He is always rushing 3 times for -1 yards a pop!
The best drive in the NFL is the 5 minute drill on defense when you are up big.
The Redskins and Falcons failed miserably at this in week 11 and the Patriots are the best.
It's also hard to modify DVOA for moron coaches who run 3 times into the line when up 3 or 7 or 8 when it is impossible to run out the clock without a first down.
Belichick and Brady are the best ever at the chess of football and that is very hard to quantify. Think of that classic Brian Westbrook play where he fell down at the 2 instead of scoring. I am sure the Eagles lost DVOA for that, but their win probability went to 100%.
#18 by Will Allen // Nov 21, 2017 - 9:36pm
44 starters, not including special teams, which also have significant influence on the outcome. It is the most complex team sport, by a very wide margin, making it exceedingly, extraordinarily, difficult to model. Which is why I enjoy observing the attempt to do so.
#76 by MC2 // Nov 27, 2017 - 4:45am
That Westbrook play might be the most overrated play I have seen in 30 years of being a fan of the NFL. Yes, it increased their WP to 100%. But if he had just scored the TD, their WP would have increased to 99%, or 98%, at the very least. So, Westbrook's "brilliant" decision increased their WP by a whopping 1 or 2%. Well, whoop dee doo!!!
#79 by MC2 // Dec 07, 2017 - 9:55am
So, every play that increases the WP by 1-2% should be talked about for the next ten years (literally -- the Westbrook play was in 2007)? Westbrook himself probably had at least 5 plays in that game that increased their WP by more than the kneel down. Why not talk about those plays for the next 10 years?
#37 by RickD // Nov 22, 2017 - 12:09am
Well that's certainly an irrelevant response. It suggests that my criticism was based on something it was not based on, while ignoring the totality of what it was based on.
"Yet they are still near the bottom of the league in defensive DVOA, even after giving up just 12.5 points per game for the last six games. New England drops a place back to 31st this week, although there's just a rounding error difference between the Patriots and No. 30 Miami."
This is how I inferred that the Patriots did not have a good defensive DVOA score for this week's game. And while you do not publish the week's rating, it certainly is something you calculated. And average in with the scores for the other weeks.
If my assumption was incorrect, you could simply publish this week's DVOA. Weekly DVOA are published fairly often here, though not always.
#44 by blan // Nov 22, 2017 - 3:00am
I think you can estimate the DVOA for this week's game as:
(This week's DVOA)*(Number of games) - (Last week's DVOA)*(Number of games - 1)
This estimate makes a number of assumptions that I think are close enough to true to give a reasonable estimate. The assumptions include: 1. Each game has roughly the same number of plays, 2. The DVOA for each play is combined via an arithmetic mean, and 3. Opponents adjustment changes are not a major factor in DVOA change for this week.
Using this estimate, I get -1.6% for New England's defensive DVOA against Oakland.
#45 by Vincent Verhei // Nov 22, 2017 - 5:40am
Perhaps we can interest you in a Premium Subscription? You can check EVERY team's DVOA, in EVERY regular season, in EVERY week, going back to 1989. More info available here:
#41 by Alternator // Nov 22, 2017 - 12:40am
I think that, at some point, we just have to make this conclusion:
Most coaches are stupid, incompetent nitwits who fail to realize that defense SHOULD change based on the situation. Given this failure, if their team is giving up gobs of yards, it's not because they're trading yards for time - it's because the defense is just playing poorly. When somebody DOES come along and realize that, yes, giving up a touchdown is fine if it chewed up five minutes, it makes them look bad because the raw stats 'look' like a team whose defense is falling apart.
There's also a second order consideration: trading yards for time is only a wise strategy when you can count on the offense to help burn clock. If your QB is Eli or Bortles, the judgement is much different than when your QB is Brady.
tl;dr: Garbage (coaching) in, garbage (ratings) out.
#26 by Hextall_27 // Nov 21, 2017 - 10:48pm
Except something like Ravens at Jaguars should be a pickem.
Bortles against that defense? How does 6-6 heading to overtime sound?
Also, the Chiefs just lost to the Giants, so I don't see big money laying 10 points on KC vs. Tennessee or San Diego.
Now if you want to talk about those winners going on the road to face a rested Patriots or Steelers team? 12-15 point favorites?
#32 by jidsardi // Nov 21, 2017 - 11:21pm
Ravens vs Jaguars would be a repeat of the London game, to a less extreme degree. Having Brandon Williams available would help depress the Jaguars' running game, but I think they'd at least be able to score. I have no idea how Baltimore's offense is supposed to move the ball.
#12 by Willsy // Nov 21, 2017 - 9:09pm
Thanks Will saw that as well. Patrick Elflein would be miffed I think.
Is it just me or are the NFC North games this year been better than the rest of the NFL in terms of entertainment value and plot lines?
The situation in GB has a Greek Tragedy feel to it. Are Rogers and Hundley just Eteocles and Polynices playing in the North? One is dead and the other soon might be. The loss on the weekend was quite vivid having watched them be so good for such a long time. If you ever needed a reason to overpay QBs the NFC Central is a great example. Prior to Etocles (Rogers) demise they looked good and are still in the top 16 of DVOA. However once Eteocles took over... Detroit was accused of overpaying their guy Stafford, really?
Looking at the DVOA numbers the North seems to have a solid collection of results. Rogers 9.2, Stafford 14.5, Keenum 31.3, Bradford 45.1. Hundley has been awful at -26.0 but Trubisky at -9.3 hasn't been a total disaster. His DVOA is the same as a few people earning a lot more than him.
Watching every game this year it struck me how much I had enjoyed the North games and when you look at the QB play you can see why. Both teams have a chance of victory and as a fan that is what you want from a contest.
Watching the Titan's Mariota last week, TB with Winston, Prescott, Eli, QBs who was know can play well are not producing this year and seem to be badly affecting the quality of the games. While Wilson has a mediocre DVOA if it had an escape from death factor in the model he might me ahead of Brady.
Thinking a good idea for a new Greek tragedy might be the Browns. The D with its middle of the pack DVOA could be the Chorus. The population of Cleveland can elect a representative to play Medea who will then kill the entire front office. Then she can give the coaching staff and O some suggestions as to how to play proper football through the provision of a magical elixir.
So the deux ex machina, after the recovery of the Browns as a functioning team, will be the lifting out of Medea by a Trump helicopter and a full Presidential pardon.
That will bring the crowds back!
#19 by DezBailey // Nov 21, 2017 - 9:42pm
Week 11 BES Rankings went out earlier today - http://besreport.com/week-11-bes-rankings-2017/
as was updated in the Week 10 DVOA Ratings comments - http://www.footballoutsiders.com/dvoa-ratings/2017/week-10-dvoa-ratings#comment-1061353
The BES and DVOA concur on the Eagles, Saints and Steelers at No. 1, 2 and 3 respectively but the Saints FELL to No. 2 in the BES after a four-week reign. Things get interesting beyond the top 3.
For example...DVOA/Weighted DVOA has the Chiefs at 10...Chargers at 16. They swap places in the BES which surprises me because Weighted DVOA and the BES tend to agree more often than not. It'll be interesting to see how that plays out in coming weeks.
#21 by Will Allen // Nov 21, 2017 - 10:08pm
Number 1 in variance plays number 2 in variance in Detroit Thursday morning. The degenerates in the peanut gallery are wondering how frequently a number 1 has played a number 2 after 10 games, and how frequently that game has resulted in an upset, compared to all games from 11 through 16.
#27 by Hextall_27 // Nov 21, 2017 - 10:54pm
Minnesota may have the hardest remaining single game of all contenders.
They will go to Green Bay in December and a very motivated Aaron Rodgers will face them. Merry Christmas!
Also, don't count out the Lions. They are 3-0 in division with 3 road wins.
#28 by Will Allen // Nov 21, 2017 - 10:55pm
I'm expecting a close game. I really hate a Thursday morning road game after playing a very physical team, like the Rams, on the previous Sunday, even though it turned out to be an easier win than anticipated. I keep expecting a regression in o-line performance by the Vikings, and they keep surprising with consistent improvement. Maybe this week I'll be right.
#48 by JoeyHarringtonsPiano // Nov 22, 2017 - 9:12am
It will be much different game than last time.
The Lions defense is playing much worse against the run, than they were back in week 4 (they really miss Haloti Ngata and Ziggy Ansah as run defenders). The Vikings offense is also orders of magnitude better than it was back then (it seems like they were kind of shell-shocked when Dalvin Cook got injured). I also assume they won’t do silly things like run Jerick McKinnon out of the Wildcat.
On the other side of the ball, Everson Griffin will find Taylor Decker a much bigger challenge than Greg Robinson (ugh). The Vikings really beat up Stafford in that game. He was limping around and played poorly in subsequent losses to the Panthers and Saints. After he got a chance to heal up during the bye week, and the pass protection improved, he’s been on a tear.
Either way, it will be a much higher scoring game than 14-7. The Vikings should be heavily favored. I’m terrified of the Vikings o-line vs the Lions d-line, which looks like a huge mismatch right now.
#22 by nat // Nov 21, 2017 - 10:18pm
I'm glad to see Aaron stating the Patriots' Q1-3 and Q4 defensive DVOAs.
DVOA has always had a weakness with roots in its main strength: it assigns relative values to yards, first downs, TDs, and turnovers based on the idea of maximizing the value of the next score assuming clock is never an issue. From that basic idea comes things like "field position is fluid", "turnovers matter more than you think", "long passes matter less than consistently moving the chains" and so on.
The problem in the fourth quarter is that situational football sometimes dictates different relative values, and introduces other things that DVOA can't assign value to at all, like stopping the clock or playing four down strategies. Those different values are highly dependent on the teams in that game, match ups, and so on. League average data just can't cut it.
There is no good way to deal with this within DVOA. The best you can do is what Aaron just did: once in a while give the quarters split and let us readers decide for ourselves what they mean.
Me, I suspect the Patriots' Q1-3 DVOA is closer to the true measure of that defense. They played one horrendous fourth quarter this season. The rest of the time was not all that bad, was situationally adequate, or once a failure that led to a middle length field goal, and was perhaps a play away from preventing even that. I think the poor DVOA showing is due to Belichick having extreme theories about how to manage prevent defenses. It's nerve wracking to watch. But if it's holding opponents to one Hail Mary or one TD attempt from 20 yards out, perhaps the theory is right.
Others might disagree. At least this way DVOA is giving us the data we need to have the conversation.
#30 by LionInAZ // Nov 21, 2017 - 11:01pm
You have a great point about DVOA providing the data needed to decide where a team actally stands. On the other hand, the subdivision only raises the Patriots' DDVOA from 31st to 26th (modulo changes to every other team), so not a difference between a bad team and a good team.
I wonder more about how DVOA handles the synergy between offense and defense. As an example, we just saw the Ravens DVOA get a big boost from playing the Packers. But do they deserve that with Rodgers out of the GB lineup? It's not just that the Packer O is worse, but also that the D has to work more because offense drives aren't sustained as long, or that there are more drives to defend against?
#39 by RickD // Nov 22, 2017 - 12:14am
Generally speaking they do not make change opponent adjustments based on injuries. It is stated that they have not done so for the Marcel Dareus trade. That might make an exception for Aaron Rodgers, though.
With only 16 games in a season, I doubt we could reasonably expect the system to adjust fully to such a major personnel change. Of course this is where Weighted DVOA comes in handy.
#50 by nat // Nov 22, 2017 - 9:45am
My point wasn't to buff the Patriots' defensive DVOA. They are pretty bad this year.
My point was to highlight how looking at Q1-3 DVOA vs Q4 DVOA can help uncover places where DVOA misinterprets effective (or even ineffective) "prevent" defense as horribly bad play overall. The Patriots are a convenient case study because (a) a lot of people are aware of their stats, (b) they've been in prevent situations a lot this year, and (c) I tend to look at their stats closely anyway.
The core issue is that DVOA measures a team's ability to generate VOA - field position, first downs, protecting the ball, and scoring TDs; all without concern for the clock or time outs used. That's almost exactly what teams are trying to do (or defend against) most of the game. But that's NOT a good model for what teams are trying to do in the fourth quarter in many games. That means a team with a bad fourth quarter DVOA might be playing badly, or it might be focusing on more important things than VOA - such as winning.
#51 by Aaron Brooks G… // Nov 22, 2017 - 11:03am
My counterpoint would be that the Pats defense collapsed in the 4th quarter against the Chiefs, too, so they do it even in games when they aren't way out front.
You can also ask Atlanta how meaningless 4th quarter DVOA is when playing with a big lead.
#55 by nat // Nov 22, 2017 - 11:41am
No one said that ALL differences between Q1-3 and Q4 DDVOA were due to prevent defenses.
Stop being so obsessed with dissing the Patriots. This isn't one of those threads. It's about how Q1-3 vs Q4 DDVOA can be used to detect cases of DVOA misinterpreting having "prevent defense" goals as being bad at "VOA" goals.
Of course, you can have "prevent defense" goals and be bad at them. You can have "prevent defense" goals when you shouldn't have them yet. You can be confused about your goals on offense or defense. None of that changes the issue being discussed here: once your proper goals are not "Maximize Your VOA" then DVOA becomes less useful in judging your team's play or predicting future play.
#58 by Aaron Brooks G… // Nov 22, 2017 - 12:38pm
That's fine. But we haven't proven that what the Pats do in the 4th is playing prevent in a manner DVOA dislikes. Because they've had the same profile in their losses, as well. It's possible they just play poorly in the 4th. That also fits their narrative.
#73 by LionInAZ // Nov 23, 2017 - 8:03pm
I've already agreed with about the statistical value of your point, and I don't think for one moment that you personally are trying to 'puff up the Patriots DVOA'. But the main reason for this thread is that ( the usual bunch)
are whining that the Patriots D is undervalued simply because they've stymied three underperforming offenses.
By contrast, lets talk about the Eagles and the Rams, the two teams ahead of the Patriots in average score lead. They don't suck at D like the Patriots do. Patriots fans should just admit the suckiness of their defense instead of trying to concoct conspiracy excuses.
#52 by dmstorm22 // Nov 22, 2017 - 11:09am
I remember there was a really nice discussion in the 2015 playoffs about this when Carolina took a 31-0 lead against the Seahawks, completely dominating them up to that point.
From then on the Seahawks went on a 24-0 run. By the end of the game, the Seahawks had a fairly decent sized edge in total DVOA, despite far fewer turnovers, mainly because they had a far better Y/P, and the Panthers didn't really do much on offense.
However, it was pointed out the Panthers win probably never dropped below 90%, they got at least one first down on most of their drives after taking a 31-0 lead, and the real opponent at that time became the clock.
Sure, we can say all the same things about Atlanta in last year's Super Bowl, but those are such uncommon events.
I really enjoyed that discussion abut the Panthers/Seahawks game - definitely made me think of how objectives change when you take a huge lead.
#74 by leviramsey // Nov 24, 2017 - 4:04pm
You hit on the fundamental issue with VOA (by extension DVOA): the "Over Average" part.
To a first approximation, VOA is basically computed by determining a success score for each play, bucketing plays based on situations, and comparing the play's success score to the average success score for the bucket (it seems likely that if a play gets half the success score of the average, then it goes in as a -50% play).
This approach works well when there's general agreement among the coaches and players in a situation about which results of a play should be considered more successful for the offense. Here performance relative to the average will be pretty meaningful. The more disagreement there is, though, the less useful a comparison to the average is (it captures more of a value difference). Most of the buckets in turn are so large (in order to get reasonable samples) that you can't adjust the weightings to account for phenomena in those situations without throwing out a lot that is predictive.
But if you judge plays against more than the average for a bucket, you can get more accurate VOA. Taking how dispersed the success values in a bucket are into account for instance, will capture situations where there's less agreement on what constitutes a good outcome and effectively deweight them automatically. For example, if a play resulted in half the average success score, but the standard deviation of success scores was 100% of the average, then the play would contribute -0.5 VVOA (Variance-adjusted VOA).
Aaron also alludes to a similar problem in taking injuries into account. Teams that face the Packers this year (where it looks likely that about half the games will be Rodgers Packers and half will be Hundley Packers) will effectively be judged as if they played a half against Rodgers and a half against Hundley, even though for everyone except the Vikings, they only faced either Rodgers or Hundley. This effect can be adjusted for by not looking as much at the average Packers VOA when adjusting for playing the Packers, but by also considering how variable the Packers' VOAs have been on offense and defense. For a team with a lot of variance (whether by injuries or coaching changes or whatever), the effect is to acknowledge in the statistic that you can't tell that much from how they did against that team.
#24 by Willsy // Nov 21, 2017 - 10:33pm
Is it correct to say the following;
The LARMs O played poorly versus the Vikes. But the Vikes D is good so they don't get penalised as much in terms of DVOA?
The Vikes D played well versus the LARMs. Since the LARMs O is good they get a bigger boost in terms of DVOA?
Nat that is an interesting point you make. Again the Vikes game had the same thing but did it inversely; 1st qtr the LARMs march down the field and score a TD. Then it was radio silence the rest of the game. So the qtr splits tell a different picture.
We use lots of different models at work to make a single decision and DVOA is just like our models, a single number that explains part of the real world.
Recalling Bill Walsh's axiom that in the 4th qtr you want a strong d line to chase the QB. While I don't have any stats on hand to support that inductively that seems a sound idea. But to your point that could be very situational i.e. if a particular D line is really good in 4th qtrs versus others then that might be more valuable than what DVOA picks up in terms of aggregate value. The point being as you say is that averages don't recognize situations properly. DVOA is based on large number sets versus small sample situational analysis.
#38 by Cythammer // Nov 22, 2017 - 12:10am
What's the most wins ever by a team that finished last in DVOA? Just wondering because the Dolphins at 4-6 have a merely mediocre, rather than terrible record. If they get a few more lucky wins we could have the prospect of the worst team in the league having nowhere close to the worst record.
#40 by junglejoe_lv // Nov 22, 2017 - 12:26am
I still think it would be very helpful to have a last 4 week DVOA for all teams that you guys provide to help vett the teams suffering major injuries. It would at least help make adjustments statistically instead of arbitrarily.
#62 by ammek // Nov 22, 2017 - 3:49pm
I don't know if you can just add all the numbers together across the special teams table (including weather and 'hidden'), but if you can, the Rams are currently 58 points above average. Working on the old maxim of 'one point = about 14 yards', that pans out to more than 800 yards of credit.
Los Angeles' +30.5 'hidden' points from special teams would be a record, wouldn't it, if they enjoy just average luck for the rest of the season? That's a flunked field goal per game!
#63 by ammek // Nov 22, 2017 - 4:07pm
It's still more likely than not that one of last year's NFC playoff teams will repeat in 2017, thanks to the one yard that Detroit could not pick up against the Falcons in week three. If the Lions had scored in three attempts from the goalline, they'd have a two-game lead over Atlanta with a tiebreaker in hand, which would make them a strong favorite for a wildcard. Oh to have retained their fourth-quarter mystique from 2016!
Per the playoff odds, the AFC championship game basically offers three equal possibilities:
2 Jax vs one of those;
3 unwatchable defensive slugfest and failed completion bonanza.
#65 by BJR // Nov 22, 2017 - 4:44pm
The biggest hope for the AFC playoffs is that the Chargers can continue improving and force their way in. They are the one team 'in the hunt' with actual talent on both sides of the ball, that could round out into a genuine threat to New England/Pittsburgh. But they are, of course, the Chargers.
#75 by SuperDackMorph // Nov 26, 2017 - 5:17am
The NFC wild card race has 3 teams, in my opinion, vying for the second alot. Carolina or the Saints will take the first. The NFC South matchups should kick Atlanta out of playoff contention. But they will have a chance, as tenuous as it might be. (They still have to play the Saints twice) The Lions remaining games (Baltimore, Cinci, Tampa, Chicago, and Green Bay) have to be the most navigable of any team right now. They could finish 11-5 but a loss to Baltimore or Cinci seems more likely; in effect, 10-6 should allow them to "sneak" into the playoffs. Last season was perhaps the sneakiest at 9-7.
I like the Ravens this year. Also, they aren't push overs like the Lions in wild card games. They are 7-1 in their last 8 wild card games. It would be interesting if they end up playing Pitt a third time in that matchup. The Steelers have their number, it seems, of late. At this juncture strength of schedule becomes more predictive of where teams end up. Can't wait for today's games.