Week 13 DVOA Ratings
by Aaron Schatz
The Seattle Seahawks have finished No. 1 in DVOA for three straight seasons. Is it possible for them to make it four straight in 2015?
The idea sounded ridiculous when Seattle started the season 2-4, but the Seahawks stayed near the top of the DVOA ratings even when they were losing close games to other strong teams. The lowest Seattle has ranked all year was 12th after the 13-10 win over Detroit that was one big officiating mistake away from dropping them to 1-3. The Seahawks were back in the top 10 the following week, despite losing to Cincinnati in overtime. They've slowly climbed the ladder since, even when they narrowly lost to Carolina and Arizona. This week's dominating win over Minnesota spawned many calls of "the Seahawks are back," but according to DVOA they never really went anywhere.
With that huge win over Minnesota, the Seahawks climbed all the way up to No. 2 in overall DVOA. Right now, they trail only Cincinnati for the entire season. Even more remarkable, the Seahawks climbed ahead of the Kansas City Chiefs this week and now rank No. 1 in weighted DVOA. It helps that the Seahawks' worst single-game DVOA of the year came all the way back in Week 1 when they lost to the Rams, and that game has very little importance in the weighted DVOA formula by this point. But it's also impressive that the Seahawks have a 65.7% DVOA in the four games since their Week 9 bye.
Maybe it seems odd to have two 7-5 teams leading the league in weighted DVOA even though there's a team still undefeated at 12-0 and three teams sitting at 10-2, but that's all part and parcel of a very strange year that combines a lack of parity with what DVOA sees as a lack of historically great teams. Obviously, win-loss record says something very different about the latter, but we'll get to that in a moment.
Eight teams are far, far ahead of the rest of the league right now. In order by total DVOA, those teams are Cincinnati, Seattle, Arizona, Kansas City, New England, Carolina, Pittsburgh, and Denver. The gap between those teams and No. 9 Green Bay is 7.5% DVOA. Those same teams are also the top eight in weighted DVOA, where the gap is even larger: 12.9% DVOA separates the Broncos from the No. 9 Jets. Honestly, the true gap is even larger than that since the No. 7 and No. 8 teams in weighted DVOA are Pittsburgh and Denver -- two teams whose offensive ratings have gone up in recent weeks because of a healthy Ben Roethlisberger and a functional Brock Osweiler replacing the corpse of Peyton Manning.
Currently, our playoff odds simulation lists seven of these eight teams with at least 7.4 percent chance of winning the Super Bowl. Pittsburgh is at just 2.1 percent because there's still a reasonable possibility of the Steelers missing the playoffs. After that, the rest of the league combined wins the Super Bowl in just 5.2 percent of current simulations.
Below these eight teams, only four other teams have a DVOA above 0% right now. Do you feel like the NFL is filled with below average teams this year? Well, DVOA agrees with you. Think of that again: only 12 of 32 teams are above average for the year, although that becomes 14 teams if we look at weighted DVOA. And a look at those 12 teams show you just how big a difference there is right now between the two conferences -- or more accurately, one conference and three-fourths of the other conference. Eight of the 12 teams that are currently above average in DVOA come from the AFC North, East, and West. None of those 12 teams come from the NFC East or AFC South.
And that huge difference between this year's good divisions and bad divisions is a major reason why DVOA is still relatively unimpressed by the undefeated Carolina Panthers. You probably aren't surprised to see the Patriots dropping from second to fifth in DVOA after their upset loss to the Philadelphia Eagles. But you are probably very surprised to see the Panthers still below them; Carolina fell from fourth to sixth despite a win over New Orleans. In fact, New England's single-game DVOA against the Eagles (-21.0%) is actually higher than what the Panthers get for their win over New Orleans (-27.3%). The Panthers had three turnovers against the worst defense in the league, and averaged only 5.7 yards on first down against a defense that has allowed 6.3 yards per play on first down all season. On defense, the problem was allowing 9.2 yards per play on second downs (including a 37-yard DPI call).
Although the Panthers are undefeated, their No. 6 position in DVOA makes sense when you look closer at point differential and schedule. They have outscored opponents 373 to 243, which would normally lead to 9.0 wins rather than 12. If the season ended now, the Panthers would set a new post-merger record for the difference between wins and Pythagorean wins. Their .250 difference in winning percentage beats the 1992 Colts (.244) and the 2012 Colts (.238).
[ad placeholder 3]
There's also the issue of schedule. Right now, Carolina's past schedule ranked 29th in the NFL based on average DVOA of opponent. Carolina's schedule this year, both past and future, includes just one team with an above-average DVOA (Seattle). If you look below at the column for non-adjusted VOA, you will see that without opponent adjustments, Carolina comes out third in the league, barely behind Arizona and Cincinnati. Before the game against New Orleans, the Panthers were No. 1 in non-adjusted VOA.
The Panthers will still be favored in all of their remaining games by our playoff odds simulation, but this week's drop in DVOA is part of why we still have their odds of going 16-0 as just 25.9 percent. And their odds of winning the Super Bowl dropped too, from 23.9 percent last week to 19.9 percent this week, though we still have them as the Super Bowl favorites.
So if the Panthers aren't a historically great team by DVOA, what about the rest of the league? Actually, while there's a huge difference between the top eight teams and the rest of the NFL, the difference between the best and worst team is actually fairly compact. Cincinnati's current DVOA of 34.1% is below average among the teams that have been No. 1 after Week 13 during the DVOA era. On the other side of the league, San Francisco improved to -31.1% DVOA with a win over Chicago. Only three teams in DVOA history were in last place with a better DVOA rating than the 49ers: the 2001 Bills, last year's Jaguars, and the 1994 Oilers. Even more ridiculous is that the only team even close to the 49ers in DVOA is Cleveland. No. 30 Tennessee is currently at -18.6% DVOA, which means that over half the league is currently sitting with a DVOA somewhere between -2% and -20%. The only other year in history with only two teams below -20% at this point was 1994.
Things are even tighter between the best and worst teams if we specifically look at offense. In part because of all the injuries suffered by past offensive juggernauts such as New England and Pittsburgh, there's currently only one team with offensive DVOA over 20%: Cincinnati, which leads the league with 22.2% offensive DVOA. That's the lowest rating for a No. 1 offense (through Week 13) since the 2008 Broncos. The last-place Rams were the highest-rated offense to ever rank last through Week 12; after their loss to Arizona, they now rank third in that category... but San Francisco at -15.1% offensive DVOA is the highest-rated offense to ever rank second-to-last through Week 13. This is the first time ever that only two teams at this point either had offense above 20% or below 20%.
Defense is a little bit more spread out. The Broncos no longer are challenging for the title of best defense ever, but their current rating of -27.4% is tenth on the all-time list through Week 13. And there is one unit in the NFL this year that is challenging historical extremes, and it brings us back to why Carolina saw its DVOA for the year drop despite a win:
| WORST DEFENSIVE DVOA EVER THROUGH WEEK 13 | |||
| Year | Team | W-L | DVOA |
| 2015 | NO | 4-8 | 24.8% |
| 1996 | ATL | 2-10 | 24.3% |
| 2001 | ARI | 5-7 | 24.0% |
| 2008 | DET | 0-12 | 24.0% |
| 2013 | SD | 5-7 | 23.6% |
| 2005 | HOU | 1-11 | 23.0% |
| 2004 | STL | 6-6 | 22.9% |
| 1999 | CLE | 2-11 | 22.6% |
| 2004 | NO | 4-8 | 22.4% |
| 1999 | SF | 3-9 | 22.4% |
| 2000 | ARI | 3-9 | 22.1% |
| 2008 | STL | 2-10 | 21.8% |
* * * * *
 |
Once again in 2015, we have teamed up with EA Sports to bring Football Outsiders-branded player content to Madden 16 Ultimate Team. Each week, we'll be picking out a handful of players who starred in that week's games. Some of them will be well-known players who stood out in DVOA and DYAR. Others will be under-the-radar players who only stood out with advanced stats. We'll announce the players each Tuesday in the DVOA commentary article, and the players will be available in Madden Ultimate Team packs the following weekend. We will also tweet out images of these players from the @fboutsiders Twitter account on most Fridays. One player each week will only be available for 24 hours from the point these players enter packs on Friday.
The Football Outsiders stars for Week 13 are:
- RE Olivier Vernon, MIA (24-HOUR HERO): 2.5 sacks, 5 QB hits.
- MLB Stephone Anthony, NO: 10 combined tackles, 7 of which were for 2 yards or less, plus a forced fumble and the first defensive two-point conversion in NFL history.
- QB Carson Palmer, ARI: Led all quarterbacks in DYAR Week 13 (356 passing yards, 2 TD vs. Rams defense ranked sixth against pass by DVOA)
- RT Mike Remmers, CAR: Panthers RB had 7 carries for 82 yards on the right side, plus no sacks allowed.
- RB Darren Sproles, PHI: 83-yard punt return touchdown, plus 14 carries for 66 yards.
With Olivier Vernon as our first Miami player, we've now done at least one player from all 32 NFL teams this season.
* * * * *
All stats pages are now updated with Week 13 information including FO Premium, snap counts and playoff odds. You can also check out further commentary on our playoff odds report at ESPN Insider.
* * * * *
[ad placeholder 4]
These are the Football Outsiders team efficiency ratings through 13 weeks of 2015, measured by our proprietary Defense-adjusted Value Over Average (DVOA) system that breaks down every single play and compares a team's performance to the league average based on situation in order to determine value over average. (Explained further here.)
OFFENSE and DEFENSE DVOA are adjusted for strength of schedule and 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. As always, positive numbers represent more points so DEFENSE is better when it is NEGATIVE. 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>
| TEAM | TOTAL DVOA |
LAST WEEK |
WEIGHTED DVOA |
RANK | W-L | OFFENSE DVOA |
OFF. RANK |
DEFENSE DVOA |
DEF. RANK |
S.T. DVOA |
S.T. RANK |
|
| 1 | CIN | 34.1% | 1 | 32.1% | 3 | 10-2 | 22.2% | 1 | -8.1% | 8 | 3.8% | 4 |
| 2 | SEA | 31.0% | 5 | 37.8% | 1 | 7-5 | 16.6% | 5 | -10.9% | 7 | 3.5% | 5 |
| 3 | ARI | 30.2% | 6 | 25.5% | 5 | 10-2 | 16.8% | 4 | -14.6% | 3 | -1.1% | 19 |
| 4 | KC | 28.5% | 3 | 34.5% | 2 | 7-5 | 13.1% | 6 | -11.3% | 5 | 4.2% | 3 |
| 5 | NE | 26.0% | 2 | 22.3% | 6 | 10-2 | 19.8% | 2 | -2.8% | 12 | 3.4% | 6 |
| 6 | CAR | 23.2% | 4 | 25.7% | 4 | 12-0 | 5.2% | 11 | -21.6% | 2 | -3.6% | 27 |
| 7 | PIT | 19.2% | 9 | 22.0% | 7 | 7-5 | 17.7% | 3 | -2.4% | 13 | -0.9% | 18 |
| 8 | DEN | 18.2% | 7 | 19.4% | 8 | 10-2 | -11.7% | 26 | -27.4% | 1 | 2.4% | 9 |
| 9 | GB | 10.7% | 8 | 3.3% | 13 | 8-4 | 8.2% | 10 | -2.8% | 11 | -0.3% | 17 |
| 10 | NYJ | 8.3% | 10 | 6.5% | 9 | 7-5 | -0.2% | 14 | -11.5% | 4 | -3.0% | 24 |
| 11 | BUF | 6.2% | 11 | 4.5% | 11 | 6-6 | 11.8% | 7 | 7.3% | 26 | 1.6% | 12 |
| 12 | OAK | 2.2% | 12 | 4.6% | 10 | 5-7 | 8.4% | 9 | 4.8% | 22 | -1.5% | 20 |
| 13 | MIN | -2.5% | 13 | 2.7% | 14 | 8-4 | -2.8% | 17 | 2.8% | 17 | 3.1% | 8 |
| 14 | TB | -3.9% | 21 | 4.3% | 12 | 6-6 | -2.0% | 15 | -0.6% | 14 | -2.5% | 23 |
| 15 | BAL | -3.9% | 14 | -5.1% | 19 | 4-8 | -5.8% | 22 | 4.5% | 21 | 6.4% | 2 |
| 16 | DET | -4.8% | 16 | -0.5% | 15 | 4-8 | -6.1% | 24 | 1.0% | 16 | 2.3% | 10 |
| TEAM | TOTAL DVOA |
LAST WEEK |
WEIGHTED DVOA |
RANK | W-L | OFFENSE DVOA |
OFF. RANK |
DEFENSE DVOA |
DEF. RANK |
S.T. DVOA |
S.T. RANK |
|
| 17 | NYG | -5.0% | 17 | -6.2% | 20 | 5-7 | -4.6% | 20 | 7.6% | 27 | 7.2% | 1 |
| 18 | WAS | -5.2% | 15 | -2.7% | 17 | 5-7 | -5.3% | 21 | 3.1% | 19 | 3.2% | 7 |
| 19 | HOU | -5.2% | 22 | -1.5% | 16 | 6-6 | -3.0% | 18 | -3.1% | 10 | -5.3% | 32 |
| 20 | PHI | -6.9% | 24 | -7.6% | 21 | 5-7 | -12.0% | 27 | -4.2% | 9 | 0.9% | 15 |
| 21 | CHI | -8.1% | 19 | -4.4% | 18 | 5-7 | 2.3% | 12 | 6.2% | 24 | -4.2% | 30 |
| 22 | STL | -9.8% | 18 | -14.5% | 26 | 4-8 | -22.7% | 32 | -11.1% | 6 | 1.8% | 11 |
| 23 | ATL | -11.6% | 23 | -13.8% | 24 | 6-6 | -4.4% | 19 | 5.0% | 23 | -2.2% | 21 |
| 24 | IND | -12.8% | 20 | -11.7% | 22 | 6-6 | -12.6% | 28 | 0.5% | 15 | 0.4% | 16 |
| 25 | JAC | -13.5% | 25 | -14.6% | 27 | 4-8 | -2.3% | 16 | 8.2% | 28 | -3.0% | 25 |
| 26 | DAL | -15.3% | 29 | -14.2% | 25 | 4-8 | -13.7% | 30 | 2.9% | 18 | 1.3% | 13 |
| 27 | MIA | -15.4% | 26 | -13.8% | 23 | 5-7 | -5.8% | 23 | 6.3% | 25 | -3.2% | 26 |
| 28 | NO | -18.2% | 30 | -18.1% | 28 | 4-8 | 8.8% | 8 | 24.8% | 32 | -2.3% | 22 |
| 29 | SD | -18.3% | 27 | -21.5% | 29 | 3-9 | 1.6% | 13 | 15.0% | 31 | -4.9% | 31 |
| 30 | TEN | -18.6% | 28 | -25.4% | 31 | 3-9 | -11.5% | 25 | 3.2% | 20 | -3.9% | 28 |
| 31 | CLE | -24.7% | 31 | -24.5% | 30 | 2-10 | -13.5% | 29 | 12.2% | 30 | 1.0% | 14 |
| 32 | SF | -31.1% | 32 | -31.4% | 32 | 4-8 | -15.1% | 31 | 11.9% | 29 | -4.0% | 29 |
- 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).
| TEAM | TOTAL DVOA |
W-L | NON-ADJ TOT VOA |
ESTIM. WINS |
RANK | PAST SCHED |
RANK | FUTURE SCHED |
RANK | VAR. | RANK | |
| 1 | CIN | 34.1% | 10-2 | 29.8% | 9.8 | 1 | 2.6% | 10 | 0.6% | 14 | 9.2% | 9 |
| 2 | SEA | 31.0% | 7-5 | 26.8% | 8.8 | 5 | 1.2% | 14 | -2.0% | 18 | 12.7% | 18 |
| 3 | ARI | 30.2% | 10-2 | 30.6% | 9.1 | 3 | -4.8% | 26 | 8.1% | 6 | 17.3% | 23 |
| 4 | KC | 28.5% | 7-5 | 24.6% | 8.8 | 4 | 5.8% | 5 | -11.1% | 29 | 20.8% | 31 |
| 5 | NE | 26.0% | 10-2 | 26.9% | 9.1 | 2 | -1.3% | 21 | -7.7% | 26 | 6.6% | 5 |
| 6 | CAR | 23.2% | 12-0 | 29.3% | 8.5 | 6 | -6.3% | 29 | -8.0% | 27 | 9.0% | 8 |
| 7 | PIT | 19.2% | 7-5 | 18.8% | 7.6 | 8 | 4.3% | 8 | 5.9% | 9 | 11.6% | 14 |
| 8 | DEN | 18.2% | 10-2 | 22.9% | 8.0 | 7 | 1.8% | 13 | 9.3% | 4 | 14.1% | 20 |
| 9 | GB | 10.7% | 8-4 | 12.1% | 7.4 | 9 | 1.1% | 15 | 3.7% | 13 | 19.1% | 26 |
| 10 | NYJ | 8.3% | 7-5 | 15.9% | 7.2 | 10 | -5.8% | 27 | -0.4% | 16 | 8.4% | 6 |
| 11 | BUF | 6.2% | 6-6 | 7.9% | 6.3 | 12 | 3.1% | 9 | -4.8% | 24 | 11.6% | 13 |
| 12 | OAK | 2.2% | 5-7 | 3.8% | 5.9 | 14 | 2.3% | 12 | 9.8% | 3 | 11.9% | 15 |
| 13 | MIN | -2.5% | 8-4 | -0.3% | 6.4 | 11 | 0.2% | 17 | 7.0% | 7 | 17.4% | 25 |
| 14 | TB | -3.9% | 6-6 | -0.2% | 5.6 | 16 | -8.4% | 31 | -3.2% | 21 | 19.9% | 28 |
| 15 | BAL | -3.9% | 4-8 | 1.5% | 5.5 | 17 | -2.8% | 23 | 28.2% | 1 | 2.5% | 1 |
| 16 | DET | -4.8% | 4-8 | -13.2% | 5.1 | 20 | 7.8% | 2 | -16.8% | 32 | 20.6% | 30 |
| TEAM | TOTAL DVOA |
W-L | NON-ADJ TOT VOA |
ESTIM. WINS |
RANK | PAST SCHED |
RANK | FUTURE SCHED |
RANK | VAR. | RANK | |
| 17 | NYG | -5.0% | 5-7 | -0.6% | 5.8 | 15 | -6.0% | 28 | -0.4% | 15 | 11.9% | 16 |
| 18 | WAS | -5.2% | 5-7 | -5.9% | 5.3 | 19 | -2.8% | 24 | -6.0% | 25 | 15.8% | 21 |
| 19 | HOU | -5.2% | 6-6 | -5.6% | 5.9 | 13 | 0.5% | 16 | -4.7% | 23 | 16.9% | 22 |
| 20 | PHI | -6.9% | 5-7 | -6.6% | 5.5 | 18 | -3.1% | 25 | 6.5% | 8 | 20.4% | 29 |
| 21 | CHI | -8.1% | 5-7 | -9.8% | 4.8 | 22 | 5.4% | 6 | -4.1% | 22 | 10.6% | 12 |
| 22 | STL | -9.8% | 4-8 | -15.5% | 5.1 | 21 | 6.7% | 4 | -2.2% | 19 | 12.5% | 17 |
| 23 | ATL | -11.6% | 6-6 | -1.3% | 4.7 | 23 | -10.7% | 32 | 3.7% | 12 | 4.8% | 3 |
| 24 | IND | -12.8% | 6-6 | -17.7% | 4.0 | 28 | 2.5% | 11 | -13.2% | 31 | 9.0% | 7 |
| 25 | JAC | -13.5% | 4-8 | -11.4% | 4.2 | 27 | -2.7% | 22 | -11.9% | 30 | 6.1% | 4 |
| 26 | DAL | -15.3% | 4-8 | -15.5% | 3.6 | 30 | 0.2% | 18 | 5.0% | 10 | 4.7% | 2 |
| 27 | MIA | -15.4% | 5-7 | -14.3% | 4.5 | 24 | -1.1% | 20 | -2.5% | 20 | 23.7% | 32 |
| 28 | NO | -18.2% | 4-8 | -16.2% | 3.8 | 29 | -0.6% | 19 | -8.4% | 28 | 13.2% | 19 |
| 29 | SD | -18.3% | 3-9 | -18.1% | 4.5 | 25 | 4.6% | 7 | 8.4% | 5 | 9.5% | 10 |
| 30 | TEN | -18.6% | 3-9 | -16.3% | 4.2 | 26 | -7.3% | 30 | 4.1% | 11 | 19.2% | 27 |
| 31 | CLE | -24.7% | 2-10 | -30.8% | 3.1 | 31 | 7.7% | 3 | 11.9% | 2 | 9.5% | 11 |
| 32 | SF | -31.1% | 4-8 | -33.6% | 3.0 | 32 | 9.3% | 1 | -1.3% | 17 | 17.4% | 24 |
Comments
114 comments, Last at 15 Dec 2015, 8:04pm
#1 by Perfundle // Dec 08, 2015 - 8:08pm
I had always felt that this season was extremely top-heavy, and the numbers confirm it. There's a stat called skewness which measures how top-heavy (positive value) or bottom-heavy (negative value) a distribution is, with 0 representing a symmetrical distribution.
https://i.imgur.com/hFR9xAE.png
The skewness of this season's DVOA's values is 0.4784, the highest in DVOA's history. Apart from 1995's 0.4554, no other season has been above 0.27. On the other hand, there have been 9 seasons below -0.27, including three below -0.6, in 1999, 2008 and 2009.
#17 by ammek // Dec 09, 2015 - 5:17am
That's very interesting. I don't know how skewing works, but the striking thing about 1995 is that there were three really strong teams. Exclude DVOA from games against Dallas, Kansas City and San Francisco, and I expect there would have been an even greater cluster around below-averageness.
#82 by Perfundle // Dec 09, 2015 - 10:40pm
Actually, if you remove those three teams, along with Pittsburgh, and normalized DVOA for the remaining teams, you'd end up with almost no skew at all.
Here's a visual representation, using 1993 (negative skew, bottom-heavy), 1995 (positive skew, top-heavy) and 2001 (no skew).
http://i.imgur.com/P8k1uOy.png
If you imagine a line connecting the highest and lowest members of a group, skewness is a measure of how many of the other members lie on one side of the line. For 1993 almost all of the teams lie above the line, while for 1995 almost all of them are below. You can also see that if you remove the first four data points of 1995 or the last six data points of 1993 the remaining teams form nearly a straight line, which is why those years are top-heavy and bottom-heavy, respectively.
#2 by jacobk // Dec 08, 2015 - 8:35pm
I get the impression that DVOA likes it better when you beat the hell out of a bad to mediocre team than when you play a great team evenly. So the next few weeks set up nicely to give the Seahawks a shot at closing the gap with Cincinnati. It's starting to feel a lot like 2012.
#4 by panthersnbraves // Dec 08, 2015 - 8:44pm
if I recall correctly, it likes consistency, so the Panthers boom/bust of throwing deep and spinning the wheel on "will Ginn catch it" is not looked upon favorably, but when they run well and grind out nice long drives, they get rewarded.
#6 by Perfundle // Dec 08, 2015 - 9:00pm
It depends on how you beat the hell out of them. For instance, DVOA liked Pittsburgh and Seattle's wins much more than Cincinnati's, despite all three being 30-point victories. That's partly because Cleveland is a much worse team, but also because Cincinnati didn't nearly dominate as much as the other two. They only had a yards-per-play differential of 2.1, compared to 3.2 for Pittsburgh and 4.0 for Seattle.
#112 by White Rose Duelist // Dec 11, 2015 - 11:50am
If you're crushing a team, you are getting a lot of positive plays, and the adjustment only affects that so much.
This is related to FO's research into Guts and Stomps from a decade ago.
http://www.footballoutsiders.com/stat-analysis/2005/fo-fox-guts-and-stomps
#7 by DezBailey // Dec 08, 2015 - 9:34pm
Week 13 BES Rankings came out earlier - http://besreport.com/week-13-bes-rankings/
Still have the Chiefs at #1 and the Panthers #2. Seahawks at #6 and Cardinals #4. Patriots fell to #5 and Bengals to #10. Just odd compared to DVOA but the BES looks at some different criteria and looks at teams' more recent body of work. Saints were #12 in the BES last week before taking the Panthers to the brink. That NFC South is somethin' else...even the Bucs are a problem.
Interesting to see the Packers 13th in Weighted DVOA. They were 26th in the BES last week, 24th this week after narrowly escaping the Lions. The BES seemed more impressed with the Lions as they moved up from 20 to 13...not far off from their 16th placing in DVOA.
Good stuff!
#8 by Will Allen // Dec 08, 2015 - 9:46pm
With Joseph, Barr, and Smith, the Vikings' best, respectively, lineman, linebacker, and db this year, all looking like they won't play Thursday in Phoenix, I'm more convinced than ever that the smart course of action, if maximizing the chance of getting to the playoffs is the overriding goal, would be run the game like a 4th preseason contest. That is, if you could do it without losing your locker room, or the league going crazy. The latter I don't care about, and I don't know enough to judge the former.
The Vikings have Phoenix in about 48 hours, then the Bears on 10 days rest at home, then the Giants at home, then the Packers. They can't function offensively when they play a good defense, unless their own defense keeps things low scoring and close. They can't do that with those three guys off the field, especially on the road. They'd be better off resting as many guys as possible, including Bridgewater and Peterson, and trying to be as healthy as possible for the two critical home games. Hell, even if they get crushed on Thursday, which is likely, if they beat the Bears and Giants, they'll be, at worst, playing the Packers for the division title in Lambeau, because 3-1 with a win over the Packers in the last game gives them the division tiebreaker. No, they very, very, likely won't win that game, but if you had told me in August, especially after Loadholdt and Sullivan went down, that they'd play for the division in week 17, I would have said ther was no better than a 10% chance of that happening.
C'mon Zimmer, Dare to Suck in the Desert, in two nights!
#9 by jacobk // Dec 08, 2015 - 10:13pm
This isn't McNesse State vs. Alabama. The Eagles just beat the Patriots last week. The Cardinals lost to the Rams this year. Play your best guys and you at least have a puncher's chance. Peterson goes off, the Cards suffer a rash of turnovers, who knows what might happen?
The Vikings still have a legit shot at a first round bye. I think any time you start a season 8-4 you should be thinking of the best way to get to the Super Bowl this year, not just maximizing the chance of making the playoffs. There really isn't such a thing as a linear development path in the NFL.
#11 by Will Allen // Dec 08, 2015 - 11:21pm
What might happen is you don't get any healthier, and you end up losing your remaining home games. They have a 5.8% chance of a bye, and I think that overstates things, in that it ignores that they will be missing their 3 best defensive players in two nights, in the game they absolutely have to win in order to gain the bye. They have a better chance of losing to the Bears and Giants, if healthy, than they have of beating the Cardinals, if healthy. They aren't healthy.
Look, I understand that making the smart move may not be smart, if you factor how the players may react, so this is mostly just speculation. However, if they get down 21 points quickly once again, Zimmer really should be smart about exposing Bridewater and Peterson in particular to more hitting that can't provide any benefit.
#10 by RickD // Dec 08, 2015 - 10:36pm
It's nice to see a system that isn't so over-impressed by a 3-point win over a bad team. Going by the GUTS vs. STOMPS criteria, the teams that had a good weekend were the Seahawks, Steelers, Cardinals, and Bengals. The Pats and Packers may have peaked too early (the Falcons obviously did). I'm reluctant to count the Broncos win over a terrible Chargers team. And can we count the Chiefs win as a STOMP if they trailed after three quarters?
#13 by Will Allen // Dec 09, 2015 - 12:39am
I could see the Pats go either way; they get healthier, and re-ascend, or they stagger on, with their old qb getting hit a lot, and they decline.
Decline relative to what, however? Until I know the Broncos are going to run consistently, I don't think they'll score enough to be obvious championship caliber. The Bengals may be it, I guess.
On the other side, Arizona or Seattle seem to be of high enough quality to be an unsurprising champ, and Carolina as well, I suppose, if Newton plays like he is capable.
#19 by FlippingADollar // Dec 09, 2015 - 8:36am
They trailed after 3 quarters due to uncharacteristically high fumbles in Oakland territory. They were moving the ball but the fumbles just killed drives. Maclin fumbled on the 43 and Kelce fumbled on 26.
On a side note, Charles Woodson is amazing.
#12 by shoutingloudly // Dec 08, 2015 - 11:37pm
I have an idea for an FO Playoff Odds Report-based measure of a division's crapulence. It is:
Sum of all division teams' odds of the 1-3 seeds, plus sum of all division teams' odds of a wild card berth.
It only works when there are still several weeks left in the season (we need some remaining uncertainty), so I think "after 12 games" is about its max value. If there are four weeks left, and your division is almost eliminated from the top three seeds AND almost definitely not getting a wild card berth, it's probably pretty crappy.
For example:
AFC South: 0.4% total odds that division champ is 3 seed or higher, plus 1% total odds of a wild card team. Pretty bad.
NFC East: 0.9% total odds of even a 3 seed, plus 1% total WC odds. And odds-on favorite division winner Philly is slotted to win the division at 7-9. Positively stinky.
Which is to say: E A G L E S, Eagles!
#87 by shoutingloudly // Dec 10, 2015 - 1:22am
Oh, good point. I hadn't noticed the overlap there. Crazy coincidence! You're right that two extra losses per team in a division would put downward pressure on a division's ability to look good...
Even if we were to give about half of those back out as victories (one more win per team), though, those divisions are still playing roughly .500 ball — clearly headed for a 4 seed and probably no wild card.
#14 by GoDog // Dec 09, 2015 - 1:44am
The FiveThirtyEight.com site using their ELO ratings has the same line of demarcation of the top eight (maybe nine) teams shown by the DVOA ratings. The same teams are in place except the ELO ratings do include Green Bay. I'm not too familiar with the ELO ratings in regards to favoring recent play over early season results which may be why the DVOA has Green Bay out of the top echelon and ELO doesn't. Regardless, it is clear this is a very skewy season.
#15 by Eleutheria // Dec 09, 2015 - 3:44am
I'm gonna preface this by saying that I love ELO, since the rest of this comment sounds very critical.
ELO isn't an advance stat or anything. It simply adjusts each rating based after each game based on the difference between expected result and actual result. (if a team does better then expected it moves up, if it does worse it moves down, simple as that.)
The reason Green Bay is still in the top 8 is becasue not enough time has passed since their big win streak to start the season. Going into week 8 the Packers had an ELO of 1685 compared to the Steelers 1582, and the games since then haven't been significant enough to result in a 103 swing in ELO.
ELO only cares about current ELO and most recent result. So for example, while most advanced statistical models are now rewarding the Vikings for their win against the chiefs, ELO isn't making said adjustments, the fact is the Chiefs were a bad team going into that game, so the Vikings aren't awarded for beating them.
#34 by TGT // Dec 09, 2015 - 11:25am
ELO was created for sports where individual play toward winning is hard to measure. Think Soccer: 80% possession isn't helpful if you can't create good chances or finish them. And what even is a "chance"? It's all subjective.
For those sports, it's the best tool available. For sports where individual play can be measured, it's a clear downgrade. I have no idea why 538 thought that ELO was a good idea for football. They wanted a Football statistic, and the useful predictors were already taken by other stats, or too time consuming?
#36 by mehllageman56 // Dec 09, 2015 - 11:35am
Really not impressed with 538's work in sports; their worth in the political arena is obvious and huge, so the connection with ESPN is nice for me on that level, but I really miss Grantland. Barnwell's articles have been using too much of ESPN's foolish stats (after the Denver loss, they still had the Pats as the Super Bowl favorite, by a wide margin), but I still prefer reading him to the 538 people.
#44 by bravehoptoad // Dec 09, 2015 - 12:16pm
ELO was created for sports where individual play toward winning is hard to measure.
Eh? This statement couldn't be more wrong. It was created for chess. It's impossible to have a sport where "individual play toward winning" is easier to measure.
#95 by bravehoptoad // Dec 10, 2015 - 11:07am
I don't believe so. Arpad Elo was a chess player in the early-mid 20th century. He designed the system for chess. I can't find any record that he ever picked up a tennis racket or watched a tennis game or even spent a half hour of his life thinking about tennis.
#76 by HPaddict // Dec 09, 2015 - 9:08pm
At first glance I also questioned TGT's comment because I agree that disentangling an individual's contributions to the performance of their football team is quite difficult. However, perhaps "individual play toward winning" did not, as you interpreted it, mean "the play of individuals" but something more similar to "a discretization into individual plays". That is, the chunking of a game into sub-portions that are treated as essentially distinct from one another.
I am not familiar with the analytic community surrounding either soccer or chess but there may well be substantial difficulties in the defining of a meaningful, and simple, discretization of individual games into distinct plays. The discretization of a soccer game probably would rely on the nebulous separation of the game into "chances" that are likely to depend on team strategy. While at first glance the turn-based play of chess provides a simple discretization, my limited understanding of chess strategy suggests that the analysis of moves outside the context of the full games. In comparison, both basketball, for which there are shots and scoring plays, and football, where there is a choice of plays, first downs or drives, are vastly simply to divide into units palatable to analysis. Different problems for different games.
I think that the use of ELO in the evaluation of drives, rather than whole games, could provide an interesting basis for analysis.
#96 by bravehoptoad // Dec 10, 2015 - 11:12am
ELO itself isn't concerned with evaluation of discrete plays. All it's concerned with is outcome. I'm sure this is one reason 538 uses it -- it's cheap and easy.
Evaluation of drives would be difficult to fit into ELO because it would require a binary "win or lose?" assessment of drive. For that kind of thing DVOA would be vastly superior.
#100 by HPaddict // Dec 10, 2015 - 1:06pm
The ELO system used by 538 is not strictly binary. While the rating of the winning team is guaranteed to increase, and that of the losing team will decrease, ELO incorporates the score differential in the determination of the size of the change. Here is the relevant quote from 538's preseason article on ELO: "Teams gain and lose ground based on the final score of each game and how unexpected the result was in the eyes of the pregame ratings." This quasi-binary system could be adapted to score drive-by-drive rather than game-by-game.
I agree that ELO itself is not concerned with the evaluation of discrete plays (although it does require a particular choice of discretization to time the updating procedure) but it may find more applicability when used to model games that lack discretization regimes smaller than individual games. Elo likely, I would guess, makes considerable use of the central limit theorem. Useful when only a 'macroscopic' picture can be developed but games with a more 'microscopic' description are likely more amenable to deeper analysis than that theorem would imply.
#45 by RoninX // Dec 09, 2015 - 12:21pm
Football is a sport where individual play towards winning is hard to measure. We're slowly getting better at it for a few positions but we are an awfully long way from being able to disentangle much of what goes on on a football field (no matter what pff may think).
ELO has major drawback and doesn't quickly respond to roster changes (e.g. injuries) but it has value in the community of metrics. If nothing else its an at a glance statistic that tells you more about where a team currently stands than W-L records (but DVOA it aint).
#59 by EricL // Dec 09, 2015 - 3:26pm
> ELO was created for sports where individual play toward winning is hard to measure.
This is almost exactly incorrect. Elo was originally created to rate chess players. It is actually at its best when there are a large number of results by an individual over time. (To the point where a chess player's rating isn't considered accurate until they've got at least 25 rated games in the system. Why it's used for football is beyond me.)
Elo has since been used in a number of different environments, but 1:1 is the preferred environment for that system.
#111 by Craigo // Dec 11, 2015 - 1:25am
Everyone else has corrected piled on the bizarro definition of ELO, but soccer analytics don't really rely on possession (which actually does have a strong correlation with winning and goal differential) or chances, but shots and their expected value.
#88 by justanothersteve // Dec 10, 2015 - 7:47am
I used to love ELO, but they really went downhill especially after Time. Rock and Roll Is King is an OK song as is Calling America, but they still don't match up to their earlier hits like Turn To Stone, Evil Woman, and their covers of Chuck Berry's Roll Over Beethoven and the Move's Do Ya.
#102 by Jerry // Dec 10, 2015 - 3:40pm
Hmmm. On the one hand, I'd never say McCartney is covering Beatles songs. OTOH, I'd probably say Wings did, if they played any back in their day, so I guess it's the difference between a complete new band and being a solo artist. So by that definition, ELO's "Do Ya" would qualify as a cover, while Lynne as a solo act wouldn't.
#55 by RickD // Dec 09, 2015 - 2:09pm
ELO is a very simple rating concept that works with any kind of competition. The way it's implemented with regard to the NFL seems to build in a lot more ratings inertia than, say, DVOA. A team that is good for a long time will continue to be able to pad its rating even by beating inferior opponents. And since the system is initialized at the beginning of the year based on last year's finish, many teams started the season with a much higher rating than the Panthers.
Green Bay is another team that benefits from having a strong record last year. Also, ELO doesn't differentiate between a narrow victory and a STOMP. So even though Green Bay hasn't looked particularly strong recently, they are still an 8-4 team.
#16 by Mr.Morden // Dec 09, 2015 - 3:52am
Shouldn't ARI-NYJ be one of the "special Super Bowl matchups" in the playoff odds? It's Todd Bowles reunion special.
Also from the playoff odds...
Car = 19.9% chance of Super Bowl win
Car = 39.9% chance of conference win
Therefore, the conditional probability of Carolina winning the Super Bowl, should it win the NFC, is 19.9 / 39.9 = 49.9%. How often does it happen that FO's "Super Bowl favorite" is projected as less than 50% to win the game, should they make it? Of course, it's going to happen only in scenarios like this, where the Super Bowl favorite is a team that's not actually in the top 4-5 or so teams by DVOA, but is only up there in Super Bowl odds by virtue of being a near lock on the #1 seed in their conference.
#18 by ammek // Dec 09, 2015 - 5:26am
Each season there are always a few counter-intuitive opponent adjustments, but this year it seems there are more than usual. For instance, Arizona's defensive DVOA gets a boost despite playing the fifth-easiest schedule, and the Packer offense gets penalized for the sixth-hardest set of opponents. Any insight into these adjustments? The Packers didn't run very successfully on the awful Bears rush defense, but there must be more to it than that.
#21 by pecxinta // Dec 09, 2015 - 9:27am
sticking with reading material pages about computer system then you need to study the data files relating to help go mainly because your your personal own will turn out to be more important after checking the article for the topic area of tourism. One of some articles linked to travel and leisure which will could fascination try to come across here an individual can likewise also view the article coupled with title a person can likewise also glimpse into this article making use of title
#25 by jtr // Dec 09, 2015 - 11:08am
I agree, pexinta, coming up short on WR depth is CLASSIC Ted Thompson. Love the Tanier riff on TT as an absent-minded mad scientist, putting together something brilliant but completely forgetting something obvious, like having enough tackles or wideouts to fill the depth chart. Green Bay's whole season and New England's last few weeks have shown that even a HOF QB needs at least one talented guy to throw to.
#40 by Will Allen // Dec 09, 2015 - 12:04pm
Look, I think Thompson is very likely a well above average personnel manager, but if at least two of the 23 teams that drafted before the Packers in 2005 had been committed to obtaining a qb at first opportunity, as is usually the case, then it is quite likely that Thompson's repuation, relative to what it is now, would have suffered greatly.
I'll note again that if Rodgers' career is winding down in 2020 or so, and the Packers are still on 3 Super Bowl appearances in the post-Lombardi era, Packers management after Ron Wolff will have significantly underperformed. If you get qb personnel capable of HOF level performance for 29 straight years (!), in a conference which has not had another dominant franchise with more than 3 appearances in 10 years, you really should get 5 or 6 trips to the last game. Now, maybe Seattle will get two or three more Super Bowl appearances in the next 5 years, beating the Packers in the playffs a couple more times. If that happens, or, of course, if the Packers do win the conference a couple times before Rodgers is done, then it is kind of a moot point. Of course, this Vikings fan thinks they'll probably get aanother HOFer at qb, this time an undrafted free agent, who'll be their starter until 2038.
I think I'm going to have to live to 150, when Blade Runner-style replicants will be playing in the league, before I see the Vikings win a championship, and only if the Vikings secretly gain access to replicant technology first. Vikings management will probably screw up downloading the expiration date of their robots, however, and they'll all cease functioning during halftime of the Super Bowl, while sitting on a 56-0 lead, resulting in the first win obtained by forfeiture. Gary Anderson's great great grandson will have written the code.
#58 by ammek // Dec 09, 2015 - 2:46pm
The front office oversight as regards WR/TE took place last offseason, not this. Last year, the depth chart was: Nelson, Cobb, Andrew Quarless, Jarrett Boykin and a cohort of rookies. Ted Thompson caught a huge break when the Packers offense had near-perfect health in 2014. James Jones would not even have been available to patch up the receiving unit if Nelson had gotten hurt.
Davante Adams and Richard Rodgers are now sophomores (and so, technically, are Abbrederris and Janis) and the Packers always need their second-year players to break through. When that does not happen, the position in question is thrown into crisis – the slow or non-development of first-rounders Nick Perry, Datone Jones and Derek Sherrod had a similar effect on the pass rush and the lines. These are problems of poor drafting and/or player development, rather than failure to anticipate holes in the roster.
The real mystery is why Eddie Lacy is having such a poor year. And, perhaps, why nobody noticed in the offseason that once-serviceable swing tackle Don Barclay is now playing at a level which is sub-Newhouse and sub-Barbre.
#91 by jtr // Dec 10, 2015 - 10:31am
It's amazing the extent to which the Nelson injury has totally exposed everyone else. You're right that this basically last year's group plus Jones (who has since had his shortcomings exposed with Oakland and NYG). They desperately needed Adams to step up this year, and instead he's been totally incapable. The rumors with Lacy have been that his conditioning has slipped this year. I guess we shouldn't really be surprised that the OT situation is weak for them, tackle had gone from a weakness to a strength in just one year, and it's never surprising when a situation like that bounces back to a weakness.
#22 by JudoPrince03 // Dec 09, 2015 - 10:28am
"Carolina fell from fourth to sixth despite a win over New Orleans. In fact, New England's single-game DVOA against the Eagles (-21.0%) is actually higher than what the Panthers get for their win over New Orleans (-27.3%)."
This is an argument why stats such as DVOA could be completely irrelevant. Philadelphia has looked absolutely abhorrent the previous several weeks. They lost to Miami then got dismantled by the Buccs, both at home, only to follow those performances by getting destroyed by the Lions. Are you telling me that the Eagles going on the road to New England is supposed to be a tougher match up than the Panthers going on the road to New Orleans? The Saints have been bad on defense this year, but at home their offense is one of the best in the league. There isn't a coach in football who would rather face the Saints on the road as opposed to the Eagles at home. Add in the fact that New England got manhandled by this Eagles team (the score would not have been that close if not for a botched onside kick) and the Pats come away with a better singe game rating than the Panthers?
Are we really supposed to take this DVOA statistic seriously?
#28 by romahane // Dec 09, 2015 - 11:17am
Well, DVOA obviously does a good job in sorting the good teams from the bad. But it is a flawed predictive tool -- like most similar stats -- in that it necessarily relies on past performance. In other words, it is data driven so you have to rely on the available data. In the case of the NFL, the sample size is always very small. A few years ago, the Chiefs looked like a juggernaut thru the first 2/3 of the season in these stats, but ended up being bounced unceremoniously from the playoffs as the pretenders they were. This year, I think the Broncos are a huge wildcard. Their measurables include a number of games in which the team was led by an injured, thus poorly performing, QB. Whether or not these values accurately predict the future performance of this team based on that biased sample is, simply, really questionable. The Panthers are also an unknown based on this stat. Their lack of strong previous opponents does not mean that the Panthers are not as good as their record indicates -- instead, it means that a good measurement of their actual quality is absent.
#50 by jtr // Dec 09, 2015 - 12:29pm
>But it is a flawed predictive tool -- like most similar stats -- in that it necessarily relies on past performance.
I'm not really sure what your point is here. Are there stats out there that predict based on future performance? Past performance is the only objective thing we've got. I don't think that DVOA would be stronger if it added a subjective component adjusting the Broncos based on someone's guess that Peyton will come back and play well. And as far as the point about the Panthers, DVOA is specifically opponent adjusted, so it is designed to account for their weak schedule. That's why their performance is measured to be very bad (-27% DVOA) when they barely beat a weak opponent.
#73 by Pen // Dec 09, 2015 - 8:04pm
I use DVOA with a 19% adjustment for Homefield advantage on the NFL Predictor. I fill out the entire season starting after week 1 results come in. I adjust every week based on the new weeks DVOA. After about week 7, I hardly have to change many results. I knew when Atlanta started fast that they'd end with a losing record because their schedule and DVOA said they would.
It's an extremely useful predictive tool, perhaps the best in existence. This week I had to adjust two outcomes: Philly beating NE and Dallas beating Washington. That's 2 our of 16. 87.5% success rate is pretty good.
#64 by usernaim250 // Dec 09, 2015 - 4:50pm
This year, DVOA told us the Chiefs and Seahawks and Ravens were better than they appeared and the Falcons worse (though anyone could look at a bunch of tight W's against mediocre teams and figure that out). Every year it seems to do the same. Far from perfect prediction, for sure, but can you name something better?
#35 by MilkmanDanimal // Dec 09, 2015 - 11:26am
Well, let's see. The Panthers let the Saints put up 38, which is the Saints' second-highest total of the year after the ludicrous 52 points in the Giants game. At home, they scored 19 on Tampa (whose pass defense has not been good, but rush defense has). They scored 31 against Atlanta's terrible defense, which is 7 less than they scored against the actually good Panthers defense. 26 vs. TEN, and 26 vs. Dallas. None of that to me screams "one of the best home offenses in the league". They were phenomenal in one game, reasonably good against a number of other teams, and pretty darn good against the Panthers this last Sunday.
Carolina is getting knocked down a bit because a bad Saints team almost beat them, and that bad Saints team played better against them than they did against a series of pretty lousy teams. That's just how opponent adjustments work; if the Patriots or Bengals had dropped 38 on the Panthers, there's not as much of a negative adjustment for DVOA because THOSE OFFENSES ARE ACTUALLY CONSISTENTLY GOOD.
#51 by Led // Dec 09, 2015 - 12:57pm
Those are good points. I will say, however, that the DVOA opponent adjustments always struck me as going a bit hinky at the extreme margins. So I think Carolina's offensive performance against NO was better than DVOA gives them credit for. Here's why. The Saints pass defense is obviously bad and very likely the worst in the league, but it's probably not really as bad as its performance over the first 11 games would suggest. The performance of any of the historically good or bad teams (or units) by DVOA is almost certainly at one of the tail ends of the expected range of performance given their actual ability level. That's what reversion to the mean is all about. So in any given game, one of those historically good or bad teams should be expected to perform based on their actual ability level rather than the outlier level of performance they had been exhibiting. Of course, the "actual ability level" is unknowable unless you could play the season multiple times and average the results. DVOA by this point in the season is a good proxy for actual ability level in most cases. But for a team like the Saints, I think it's a safe assumption their pass defense DVOA for the 11 games so far overstates their terribleness, and that over the next 11 games (if there were 11) their DVOA would be somewhat better. Tinkering with the DVOA model to deal with the occasional outlier like this would probably make it less predictive overall, so this is not a major criticism. But whenever there are massive opponent adjustments, I tend to discount them somewhat in my head.
#53 by InTheBoilerRoom // Dec 09, 2015 - 1:49pm
I have made the same mental note when viewing single game DVOA involving teams at either extreme. The Panthers, a middle-of-the-pack passing offense (by DVOA) going up against the worst passing defense in the league, subjectively seemed to have performed on the same level against NO as other teams have.
If Cincinnati (#1 passing offense by DVOA) were to play NO, how flawless would they have to play to actually have a single game passing DVOA matching their current season long passing DVOA (currently 55.6%)?
#66 by Kal // Dec 09, 2015 - 5:23pm
Probably a better argument against DVOA this season is the Seattle-Arizona game as Aaron mentioned - where despite Seattle having a sub 50% completion % and having a bunch of totally failed offensive drives, they ended up with a higher overall DVOA than Arizona did.
Part of that is that DVOA just doesn't give big plays a lot of credit. Part of it is the penalty issue, where getting 8 yards on 1st and 20 is a 'successful' play, but getting the loss of 10 to begin with isn't a big deal. And part of it is that DVOA cares more about successful plays but not necessarily strings of successful plays. A couple years back there was a NYJ/Pats game where NYJ ended up with a hugely good offensive performance despite scoring like 21 points. The reason was that they had something like 7 3 and out drives (and most of those plays were successful on those drives) and then 3 10+ play drives where each play was pretty successful. End result? A ton of successful plays, very little in the way of points or actual possession or field position.
#68 by panthersnbraves // Dec 09, 2015 - 7:26pm
That's part of why the Panthers have a lower DVOA, i think. Due to teams stacking the box/spying, combined with a meh receiving corps (outside of Olsen) and you get the Panthers throwing a LOT of deep balls. Some overthrows and some drops. So there are a lot of unsuccessful plays - boom/bust. So the opposite here - wins but low DVOA.
#72 by Perfundle // Dec 09, 2015 - 7:44pm
I don't think so. Carolina's Drive Success Rate, the percentage of down series that result in a first down or touchdown, is 7th in the league, which is higher than their offensive rank. I think the sole reason why they're as low as they are is because of strength of schedule. Sagarin's Predictor, which only looks at point differential and strength of schedule, had Carolina 7th. Pro Football Reference's SRS, which looks at the same things, also has them 7th. Meanwhile, they're 3rd in VOA, and nearly tied with Cincinnati for 2nd.
#71 by Perfundle // Dec 09, 2015 - 7:37pm
Seattle having a sub 50% completion %
Part of that is that DVOA just doesn't give big plays a lot of credit.
You're contradicting yourself. The reason that Seattle had a higher yards-per-play, and more specifically a higher yards per dropback (6.71 per Seattle, 6.68 for Arizona) despite a sub 50% completion rate was because Seattle had a higher percentage of big plays. They averaged 17.1 per completion compared to Arizona's 12.5.
As far as what DVOA cares about, I think it's very reasonable. It might be inaccurate at determining who played better in a single game, but it's better for predictive purposes, which is what it's designed for after all. An offense that can consistently generate successes on 70% of plays will be very good on average, and a game where the successful 70% are all piled up in three drives with the rest being three-and-outs is simply a fluke.
#74 by Kal // Dec 09, 2015 - 8:10pm
Seattle averaged a higher YPC because they had fewer actual completions though, yes? That's sort of the problem. Seattle didn't actually throw deep a whole lot in that game. They got a bunch of 20ish yard completions, but nothing like what Arizona got. The average ended up about the same, but the distribution is a lot more weighed at the ends for Arizona and the middle for Seattle.
My point there was that Arizona's offensive value was reduced because they got bigger plays than Seattle did.
Furthermore, Seattle got a lot of 'successful' plays but didn't do much with them, and that ended up sucking. And that's been a problem for them all year, so that is somewhat predictive.
I'm also not convinced that it's better predictively because of it. Better than what? One of the things that is interesting to me is that S+P incorporates an explosive play part of the overall value, and that the best indicator of who won a game is which team was more explosive. While marching down the field consistently works, it doesn't matter if you don't score points - and explosive plays are definitely a big deal as far as scoring points.
To clarify: the only thing that DVOA predicts if it is measuring successful plays is whether a team will get successful plays. That sounds tautalogical, but the point is that successful plays do not indicate scoring or winning the game by themselves. And furthermore, successful plays only matter when they're done in a row, consistently.
What would be more predictive? Possibly measuring something like what team gets runs of successful plays together. Or what team stops said runs. That's closer to drive-based stats like FEI does, and those are predictive in other ways. It's also possible that in the NFL the team that can get more explosive plays is also more likely to win, and while they might not be quite as predictive, some teams and some personnel lend themselves a lot more towards getting big plays compared to others. That gets into things like ALEX.
Finally, I don't see how looking at the Seattle/Arizona game could be particularly predictive of future events. Their offense scored 18 points on drives that they didn't start on the opponent's 3 yard line. Arizona doesn't have a particularly strong offense. What of that game (and Seattle's well above average DVOA) would indicate that their offense would be one of the most efficient down the road? Why is getting 8 yards on 1st and 20 indicative that you'd be able to get 4 yards on 1st and 10?
#75 by Perfundle // Dec 09, 2015 - 8:54pm
They got a bunch of 20ish yard completions, but nothing like what Arizona got. The average ended up about the same, but the distribution is a lot more weighed at the ends for Arizona and the middle for Seattle.
This is wrong in every particular. Seattle's completions were for 40, 33, 32, 30, 28, 14, 10, 10, 9, 8, 8, 8, 7 and 3 yards. Arizona's were for 35, 27, 22, 21, 20, 19, 15, 15, 14, 13, 13, 12, 12, 11, 11, 10, 10, 10, 9, 9, 8, 8, 8, 7, 7, 6, 6, 5, and 0 yards. Both teams actually had the same number of 20-yard completions (You said ish, so I'll give Arizona the 19-yard one so they lead 6-5), but Seattle's were much longer. Seattle also had five 10-yard runs to Arizona's one. Considering your entire thesis rests on the faulty belief that Arizona had more big plays than Seattle, I think it's time for you to retire it. I don't even know why you think that DVOA doesn't think that much of explosive plays, because they absolutely do. They somewhat discount extremely explosive plays, so that a 90-yard play isn't that much different from a 50-yard play, but that's it. If anything, Seattle's DVOA was so high that game because they had so many big plays.
Furthermore, Seattle got a lot of 'successful' plays but didn't do much with them, and that ended up sucking. And that's been a problem for them all year, so that is somewhat predictive.
Apparently not, considering they've been number one in offense for the last three games, precisely because they've been doing a lot more with them. It looks like the prediction that Seattle stringing together good plays only outside of the red-zone, as well as their predilection for holding calls was unsustainable and they'd eventually perform better there was the correct one. Their current level of offense isn't sustainable either, but I'm puzzled that you're using Seattle to argue for the inefficiency of DVOA when they've been validating the predictions enormously in recent weeks.
#78 by Kal // Dec 09, 2015 - 9:46pm
That's exceptionally weird about Arizona. That's not at all what I remember about the game; I remember Arizona going deep on play after play - both running and receiving. But you're right, that is the breakdown. And while it doesn't count the CJ performance, it seems like that's legit.
From DVOA's discussion in the past, it's talked about how plays longer than 20 yards are pretty much negated as far as predictive quality. An example of this is seeing Mariota's 87 yard run as being worth only 24 DYAR, essentially the maximum value for a run.
As mentioned in the DVOA breakdown for that week (can't link to it directly, it's the week 10 DVOA) - Seattle got a huge bonus on offensive DVOA, but it wasn't because of big plays. It's because of a relative large amount of very successful plays and otherwise quick outs - kind of like that NYJ game I mentioned.
The problem with having Seattle being 'predicted' to do well in the last 3 games should be pretty obvious, but let's go through it anyway. Seattle lost Marshawn Lynch and Jimmy Graham and replaced their center. There's no reasonable expectation that would indicate that because they were successful earlier with those personnel (though again, why you would call the Arizona game 'successful' is pretty weird) that they would be successful without them. Their predilection for holding calls and bad calls hasn't changed at all; they're still one of the most penalized teams in the league, still getting offensive penalties all over the place (such as the holding call that wiped out a 51 yard run or the roughness penalty on Sweezy at the goal line). They've been converting 3rd downs at a very high rate, and one of this site's common assertions is that 3rd down success is not sustainable and their success has largely come to better passing, which until the last few weeks was not how they were doing.
If you had been looking at Seattle and how they would be predicted to do by DVOA analysis this came completely out of left field. Claiming that DVOA successfully predicted this run is ridiculously fitting the result of the data. Even Aaron stated that this was likely the case in the above analysis. Heck, if you look at the same article you'll notice that Minnesota also was predicted to be looking better and on the upswing - so why did they melt down?
Another way to say it is this: what from the Arizona game and Wilson's 14/32 performance would indicate that he would have 130+ passer ratings in the following three weeks, especially after losing Jimmy Graham? What from that game would make you think that that game was actually better than how Seattle played against Minnesota (which by DVOA it was)?
My argument is mostly that I get bugged when DVOA has 'pegged' Seattle as being this awesome offense (and now it's showing) when the reasons it had pegged it as having a great offense don't actually appear to align with reality all that well. Another way to say it is this: why do you believe that Seattle's offensive performance against Arizona at home was one of the best offensive performances of the season? How does that performance relate well to what they did in the 3 weeks after? And why were those 4 weeks so much better than the previous 9, where they had a league-average DVOA?
#83 by Pen // Dec 09, 2015 - 11:12pm
I can address both Minnesota and Seattle. Minnesota didn't melt down. Using the NFL Predictor, giving home field a 19% advantage, DVOA said Minnesota would do exactly what they're doing. It's not a melt down, it's simply their schedule and opponent. Seattle is just that much better than they are. Even if Minnesota is improving, it would be unlikely they'd improve so much as to significantly alter the likely outcomes of their upcoming games.
As for Seattle, there is more happening there than DVOA or any statistical analysis can factor in. After the bye week, they went to a spread offense and a quick passing game. They knew Lynch wasn't 100% and couldn't count on a UFDA rookie to take his place. Rawls, however, was already playing lights out per DVOA and should have replaced him all along, but NOBODY is going to bench Lynch for a rookie. so since the bye, the quick passing game utilizing a spread offense and moving away from play action has immensely helped Wilson get the ball out before his oline can collapse. And Rawls has been a huge upgrade over Lynch. These two factors cancelled out the loss of Graham and Lynch. But the most important thing that explains why Seattle is doing so much better is the end of the Nowak experiment at center.
#85 by Perfundle // Dec 09, 2015 - 11:39pm
Seattle got a huge bonus on offensive DVOA, but it wasn't because of big plays. It's because of a relative large amount of very successful plays and otherwise quick outs
Well, the very successful plays are the big plays.
I'll admit I was wrong about the holding calls. I thought there considerably fewer in recent games but they haven't been. However, in regards to third-down conversion rate, the ridiculously low percentage they had wasn't sustainable either, given how good their offense was on first and second downs (outside of penalties). Neither was their red-zone TD rate, which was dead last before the bye.
Heck, if you look at the same article you'll notice that Minnesota also was predicted to be looking better and on the upswing - so why did they melt down?
Because they ran into a team even more on the upswing than they were. Seattle's defense also felt they had something to prove after getting destroyed in the previous game, so that didn't help.
Another way to say it is this: why do you believe that Seattle's offensive performance against Arizona at home was one of the best offensive performances of the season?
Because they outperformed nearly all of Arizona's other opponents on a per-play basis. Here is the yards per play Arizona has allowed in each game:
5.8 New Orleans Saints
5.4 Chicago Bears
3.1 San Francisco 49ers
6.4 St. Louis Rams
4.9 Detroit Lions
5.8 Pittsburgh Steelers
4.7 Baltimore Ravens
4.2 Cleveland Browns
6.6 Seattle Seahawks
5.3 Cincinnati Bengals
6.7 San Francisco 49ers
4.2 St. Louis Rams
And that 6.6 total isn't because of 8-yard gains on 1st-and-20 either. Seattle was actually horrible in long yardage situations, averaging a paltry 2.4 yards on their 16 plays with more than 10 yards to go. Most teams don't get many yards in those situations either, so Seattle was not particularly punished there. On the other hand, they were rewarded for their absolutely bonkers average of 8.5 yards on plays with no more than 10 yards to go, compared to Arizona's opponent average of 5.4. I don't know why you keep emphasizing that 14-32 number and ignore that A) Wilson still averaged 7.5 yards per pass, and B) the 6.4 YPC they got on the ground, the highest against Arizona this year.
#103 by Kal // Dec 10, 2015 - 5:56pm
I emphasize their poor performance because offensively it wasn't very good. Again, 20 points scored when they didn't have to start at the opponent's 3. In addition to that, they had 4 drives that scored points - and their other drives did the following:
7 plays, 8 yards
3 plays, -5 yards
1 play, safety and -20 yards
3 plays, -8 yards
2 plays, 6 yards and Int
8 plays, 5 yards
7 plays, 32 yards
Their 4 other drives looked pretty good, I guess - but the above is a pretty inept set of offensive performance. Heck, the more I look at it the more it looks a lot like that NYJ/Pats game I mentioned - where they had a whole bunch of complete failure drives with a few successful plays on those drives (like Lynch running for 10 yards on 1st and 25) and a few drives with basically nothing but successful plays.
But to me, the above does not indicate a raw offensive above average performance of 50%, which is what they received.
I emphasize the 14-32 because of a few things. One is that this means 18 plays were unsuccessful, period. That's a big deal. Another is that it is completely unlike what the Seahawks have been playing like in the last 3 weeks; saying that it predicted what is going on now makes no sense. Third because most analysts called it Wilson's worst game in his career, but DVOA looks at it as super good.
And the 6.4 YPC is kind of stupid to look at when you're talking about this game in a vacuum. Lynch running for 10 yards on 1st and 25 isn't all that successful, period, and it sure isn't the same as a 10 yard run on 1st and 10. If that's a positive thing for Seattle and is predictive of future success then the system is likely very wrong.
#106 by cardbomb // Dec 10, 2015 - 6:07pm
Couldn't have said it better myself.
I'd like to emphasize even more the "20 points scored when they didn't have to start at the opponent's 3" Minus, those turnovers in the game, which have nothing to do with the Wilson or Seattle's offense, the Cardinals win in a blowout.
#107 by formido // Dec 10, 2015 - 7:30pm
Seattle hasn't been blown out since before Wilson arrived.
I'd say Arizona was massively helped by a 10+ yard penalty being called on the offensive line in each of the first 4 drives, which helped allow Arizona to build a 19-0 lead. I wouldn't count on that as a way of predicting Seattle's future offensive success.
#110 by cardbomb // Dec 10, 2015 - 11:51pm
Seattle hasn't been blown out since before Wilson arrived.
Therefore they can't be blown out. Great argument.
I'd say Arizona was massively helped by a 10+ yard penalty being called on the offensive line in each of the first 4 drives, which helped allow Arizona to build a 19-0 lead. I wouldn't count on that as a way of predicting Seattle's future offensive success.
Yes, and Seattle was helped by a non-call that should have been a safety. That definitely helped with inaccurately boosting their DVOA. Seattle's offense looked bad that game. There's no way around it.
#84 by gomer_rs // Dec 09, 2015 - 11:29pm
Your criticism would be somewhat reasonable if Seattle hadn't done this before. In 2012 Seattle took a giant leap in offensive efficiency when they stopped treating RW as rookie QB that needed to be protected and began to open the whole playbook for him after playing against the Chicago Bears.
Seattle is undergoing a similar transition in efficiency by transitioning from a run first football team to pass first team. This transition is replacing less efficient running plays with more efficient passing plays. Less play from under center, more play from the shotgun.
What DVOA may have been telling us is that Seattle was a talented team just missing and by replacing a few inefficient plays with efficient ones they can grow quickly.
_______
I remember when they were the Sea-chickens.
#104 by Kal // Dec 10, 2015 - 5:58pm
I don't disagree with anything other than 'what DVOA was telling us'. DVOA was telling us that prior to the Arizona game Seattle was an average offense. Then Wilson having the worst game of his career meant Seattle was a good offense. Then, losing Lynch and Graham for the season, Seattle was a great offense.
I think that you're right - that Seattle has basically given up on the 'we're going to be a running team of awesomeness' and put the game in the hands of Wilson to win. They've put in better pass blockers on their line and got rid of the idiotic Nowak experiment. They've done more interesting PA pass and passing concepts. They've upped the tempo a bit too. All of those things are true.
But none of them are particularly predicted by what they did against Arizona.
#108 by gomer_rs // Dec 10, 2015 - 7:55pm
"But none of them are particularly predicted by what they did against Arizona."
This actually isn't true. The Arizona game is when they began the transition. They played about 50% run first I formation and 50% pass first shotgun, generally depending upon point differential. When they played pass first they tore through a very good defense and when they played run first they were penalty ridden and ineffective.
I have to imagine it was the stark difference in efficiency on tape that caused the change in offensive styles. Before Arizona they're the Seahawks as we knew them, during Arizona they're 50/50, after Arizona they're a spread pass first team.
_______
I remember when they were the Sea-chickens.
#60 by Thomas_beardown // Dec 09, 2015 - 3:28pm
While DVOA is adjust for opponent strength (toughness of matchup as you put it), that's not the only input, how each team performed on a per play basis is also accounted for.
While the Eagles scored a lot of points, they didn't do so through lots of successful plays that you would associate with a good football teams. Instead they scored a lot through a number of one-off fluky plays that don't indicate how strong they are or how weak their opponent is. Now, they still played pretty well, and thus the Patriots have a strongly negative DVOA (if they played like this all year -20% DVOA would make them the 30th ranked team in the league).
The Panthers did not have fluky plays go against them, they scored a lot of points, but against one of the worst defenses ever, and got a bunch of points scored on them by a good, but not fantastic offense.
#63 by usernaim250 // Dec 09, 2015 - 4:47pm
The Pats got manhandled? That tells a lot about how you "look" at football. They gave up 21 points on returns--likely at least half of what they will surrender all year, in one game. And yet possessed the ball with a chance to tie and time enough to do it. They outgained the Eagles 427-248 and beat them in just about every other meaningful stat. And for whatever you'd like to say about the Eagles D, the Saints give it up like that to almost everyone, including mediocrities like the Giants and Redskins.
#23 by Raiderjoe // Dec 09, 2015 - 10:46am
Raiders may go 9-07 and still miss playoffs. so will just rank afc teams in rootng interest.
want to see broncos, chioefs, and Pates fail in playoff. 2 will, all 3 might.
next up si steeler snad Clots.
can tolerate Texans, jets and Benagls.
nFC like cards and Pahnthers moist. then Vikes, Squirrels, Buccs, Pakckers, Gianta, eags all jumbled.
hate Seahwks. want them to lose.
#89 by Bill Walshs Ho… // Dec 10, 2015 - 10:18am
My neighbor's aunt once had snad Clots; was in the hospital for weeks. She forever drooled out of the right side of her mouth.
I also like eags all jumbled for breakfast.
My cat's breath smells like cat food.
#29 by TGT // Dec 09, 2015 - 11:18am
And they could very well end the year 15th...at 4-12. Their remaining schedule is the 1st, 3rd, 4th, and 7th ranked teams, 1st, 2nd, 3rd, and 7th by weighted DVOA, and the 7th in both is Pittsburgh, who are probably underrated due to missing Roethlisberger mid year.
#101 by Grendel13G // Dec 10, 2015 - 3:14pm
Simulations based on the behavior of lots of small, discrete pieces (agents). For football, this would probably mean simulations with representations of actual players in some way. Letting Madden play itself would be an elaborate example of an agent-based simulation, though they certainly don't have to be that faithful to real life.
#46 by johonny // Dec 09, 2015 - 12:22pm
It's depressing that Miami isn't particularly good at anything. It is not like you can say "Let's build on that top ranked defense, or at least we got good special teams. The team is in the bottom of the league in nearly everything. They have little cap space next year and what space they create is going to be needed to resign those young pass rushers. It some point Stephen Ross has to sell this mess... I hope.
#47 by Independent George // Dec 09, 2015 - 12:23pm
How does Tampa's ST come out for weeks 5-13 only?
I ask because it looks like it's their FG/XP that's dragging down their otherwise good ST play, and week 5 was when they replaced Kyle Brindza (6-12 FG, 6-8 XP) with Connor Barth (18-21 FG, 17-18 XP):
FG/XP: -14.9
Kick: 3.0
Kick Ret: 0.5
Punt: 2.8
Punt Ret: -0.7
Hidden: 2.8
Weather: 5.8
#65 by usernaim250 // Dec 09, 2015 - 5:01pm
How about adding a couple of things to playoff odds going forward?
a) Median wins to get the second wild card (i.e. do sims see 9-7, 10-6 etc.).
b) Median wins to be division champ for each division. I'm curious whether the median div. champ of the east might be at 7-9, for instance.