by Aaron Schatz
If you've been reading Football Outsiders the last couple weeks, you know the drill on this one. Super Bowl XLIX is a close matchup if you look at our standard regular-season DVOA, where the Seattle Seahawks finished No. 1 and the New England Patriots No. 4. But it's really, really close if you use our weighted DVOA ratings that include playoff performance and gradually lower the weight of games from more than two months ago. In fact, it might be the closest Super Bowl matchup ever.
... but again, it's only the closest Super Bowl matchup ever if you don't consider the first few weeks of the season. This gets to what has been a running theme this year: just how much data is appropriate to judge a team, and is it really more accurate to judge a team without including its early-season games? We've written about how surprisingly, total regular-season DVOA tends to predict playoff performance better than weighted (although also regular-season only) DVOA. At the same time, you don't want to allow tiny differences in correlations to overwhelm common sense.
There's no doubt that the New England Patriots and Seattle Seahawks are better teams now than they were in September and early October. It's not a question of just looking at the records, where the Patriots began 2-2 and the Seahawks 3-3. There are major personnel changes that help to explain the statistical improvement in the second half of the year, in particular players who either returned from injury/suspension (Jeremy Lane, Brandon Browner) or played through early injury and then improved as they got healthier (Rob Gronkowski, Bobby Wagner, Kam Chancellor). The Seahawks offense improved when they stopped desperately trying to run the entire scheme around Percy Harvin. The Patriots offense improved when they finally installed rookie Bryan Stork at center, then again when they picked up LeGarrette Blount off waivers from the Pittsburgh Steelers.
Analysis of the stat matchups in this game is made more confusing by the way those stats changed dramatically over the course of the year as the two teams improved -- or, in a handful of areas declined. The running theme of this preview is to ask "which Seahawks team are we looking at, and which Patriots team are we really looking at?"
For those who may be unfamiliar with the Football Outsiders stats, they are explained at the bottom of the page. Scroll down or click this link. Please remember that all stats represent regular season only, except for weighted DVOA and anything else specifically noted. All game charting for these two teams is now complete; any game charting data that appears with a asterisk appears courtesy of the ESPN Stats & Information Group. This preview has two different week-to-week charts for each team, one for offense and one for defense. Because defensive DVOA is opposite of offensive DVOA, the defensive charts are flipped upside-down; thus, the higher dots still represent better games.
You also might want to read the rest of our Super Bowl XLIX content from over the last couple weeks.
- I looked at Seattle and New England ranking among the league leaders in penalties, and the strangely low total of penalties called against Seattle opponents.
- Cian Fahey broke down how the Seahawks might cover Rob Gronkowski and what the Patriots' linebackers mean to this game.
- Scott Kacsmar looked at Russell Wilson's mobility and how often he improvises by charting every play of his NFL career.
- Vince Verhei looked at the common threads between New England's losses as well as Seattle's losses.
- J.J. Cooper looked at the two offensive lines and their trends in sack prevention this season.
- Scramble for the Ball had its annual Prop Bet Extravaganza.
All readers can click here for in-game discussion on our message boards. If you have FO Premium, you can click here to see all the matchup of DVOA splits for this game.
*In keeping with the theme of these two teams transforming after the early weeks of the season, this is no longer an issue for the Seahawks, as explained further here.
WHEN THE SEAHAWKS HAVE THE BALL
Seattle is one of the few NFL offenses still based around the ground game despite the pass-heavy environment of the modern NFL. The Seahawks run more than almost any other team, and they do it with remarkable efficiency that mixes steady gains with huge game-breaking highlight runs. Seattle's run offense DVOA, which incorporates both runs by the running backs and by quarterback Russell Wilson, was the fifth highest of any team since 1989.
A quick look at New England's season-long stats suggests that the Seahawks won't find it too tough to run on the Patriots. They've been an average run defense over the course of the entire season... but this is the first of many places where one of these teams has changed dramatically since the early part of the season.
From their Week 10 bye until the end of the regular season, the Patriots had the best run defense in the league according to DVOA. There's no clear change that seems to explain such a strong turnaround. It's not really an issue of linebackers, as both Jamie Collins and Donta Hightower were playing plenty even before Jerod Mayo went out for the season in Week 6. The Pats added defensive tackle Alan Branch in Week 11, but he didn't play 40 percent of the snaps in a game until Week 16. Another rotation tackle, Sealver Siliga, returned from the IR-designated list, but not until Week 14.
|Patriots Run Defense, 2014|
|Shaded cells are stats that include all runs, while the other stats are for running backs and Wildcat plays only.|
The improved run defense disappeared in the two playoff games, although that's obviously a small sample size. The Patriots allowed a combined 4.79 yards per carry and 13.7% DVOA to the Baltimore and Indianapolis running backs.
The Patriots don't allow a lot of long runs (No. 3 in the league, allowing just 0.47 Open Field Yards per carry) but they don't force a lot of short runs either. They stuff opposing running backs on just 15.7 percent of runs (28th in the NFL) while Seattle running backs are stuffed for a loss or no gain on just 17.5 percent of runs (sixth among offenses). The numbers look even worse for the Patriots in those must-convert situations. Seattle converted 81 percent of short-yardage runs, tied for the league lead, while the Patriots allowed conversions on 81 percent of these runs, the worst of any defense in the league. Like all the Pats' run defense numbers, these stats got better in the second half fo the season. The stuff percentage is particularly improved after Week 11, but the ability to stop opponents in short-yardage situations was still a problem.
Another issue is that the Seahawks run best where the Patriots' run defense is at its worst. Our ALY numbers show the Pats as strongest against runs on the left and weakest against runs up the middle or to the right. And the Seahawks running backs are strongest when running to the right, and weaker when running to the left, though they are better running "left end" than "left tackle."
There's also the type of run we're talking about here. According to ESPN Stats & Information data, the Patriots faced 38 read option plays this year, exactly the NFL average, and did allow a less-than-average 4.18 yards per carry. Cian Fahey talked about it a bit in his piece on the Patriots' linebackers last week, but 33 of those 38 plays came against two teams, the Dolphins and Jets. The Dolphins only ran one zone read in Week 1, so we're really just talking three games, plus a handful of plays by the Raiders and Bengals. There was a blown handoff by the Jets charged to Geno Smith, but only two of the other 37 plays were quarterback keepers, Ryan Tannehill runs for 15 and 6 yards in Week 15. That's 5.4 percent of the zone reads against the Pats. By comparison, Russell Wilson kept the ball on 18.8 percent of Seattle zone reads this season, not counting blown handoffs. So the Patriots' defense against the Jets and Dolphins doesn't tell us what they'll do against a better running back and a quarterback more likely to keep the ball.
Connected to the read option, in part, is the play-action game. The Seahawks used play-action on 31 percent of pass plays this year, second in the NFL, and many of those fakes were built to look like zone reads (or were actually zone reads where Wilson kept and threw instead of handing off). Seattle actually got a below-average benefit from play-action, with average yards going from 6.6 without to 7.4. That's an 0.8-yard advantage, compared to the NFL average of 1.3 yards. The Patriots faced very little play-action -- 17.7 percent of passes, 30th in the NFL -- and were average against it, allowing 7.4 yards per pass with play-action and 6.0 yards per pass without.
New England's pass defense improved at the same time the run defense did, from 10.2% DVOA in Weeks 1-9 to -8.1% DVOA after the Week 10 bye. Pass defense around the entire league improved significantly in the second half of this season, so that only took the Patriots from ranking 16th in pass defense in the first half of the year to 11th in the second half. However, there is a clear personnel change here that led to the improvement in the pass defense: the activation of Brandon Browner after an early-season suspension and a couple of games missed due to an ankle injury. More importantly, activating Browner meant the Patriots didn't need to depend on undrafted rookie Malcolm Butler or the suddenly, inexplicably terrible Alfonzo Dennard.
|New England Cornerbacks, 2014|
Darrelle Revis led the Patriots in targets mostly because Browner only played nine games, but make no mistake, he's not Revis Island anymore. He's just Revis, Very Good Cornerback. Opponents were perfectly find throwing at Revis a few times per game. His coverage stats are similar to those for Richard Sherman, with a slightly higher Success Rate but more yards allowed per pass. As you might expect, Revis moves around the field more than almost any cornerback in the league. We have him charted with 28 targets on the left side, 30 on the right. Some thought that the Patriots might revert to cornerbacks-by-sides once Browner was healthy since he had been used to always playing on one side in Seattle, but he moved around too, with 26 targets charted on the left and 18 on the right. Revis generally does cover the opponent's No. 1 receiver and I suppose that means Doug Baldwin in this game, with safety help likely rolling towards Browner covering Jermaine Kearse. Kyle Arrington is the preferred slot corner. Ryan is the utility guy, who covered for Browner early in the season, covered for Arrington late in the year, and came in when the Pats went to a four-corner, one-safety "penny" set like they did against the Colts in the AFC Championship. Butler is primarily a special teams player who will only play defense if there's an injury; Dennard lost his job early in the season and has only two charted targets after Week 7.
Seattle's passing game improved after the early part of the season, just like the New England pass defense, and again a specific personnel change is clearly part of the improvement. After Week 6, the Seahawks weren't trying to build their entire offense around 2-yard passes to Percy Harvin, and they went from 10.4% passing DVOA and 6.12 net yards per pass before the trade of Harvin to 23.8% DVOA and 6.71 net yards per pass afterwards. Harvin actually had the worst DVOA of any of the Seattle wide receivers this season.
The wide receivers are not the strongest part of the Seattle offense. Doug Baldwin is very good, as noted in this piece about catch radius before the season, but he still had just 5.4% receiving DVOA. The other starting receiver, Jermaine Kearse, was just 61st in DVOA among wideouts with at least 50 passes. Paul Richardson, the only other receiver with more than 15 targets, is out with a torn ACL.
Making up for that weakness are the running backs, who are very good as receivers, as well as tight end Luke Willson, a 2013 fifth-round pick who has broken out big in the second half of the season. Since Week 10, including the postseason, Willson has 47.6% DVOA, catching 20 of 29 targets for 362 yards and three touchdowns. Willson could have a huge game because the Patriots have had major problems covering tight ends all season, ranking 30th in the NFL. Their DVOA against tight ends improved slightly in the second half of the year, but was still poor at 13.4%, and they actually allowed a higher catch rate to opposing tight ends after Week 10 (68 percent, as opposed to 63 percent before Week 10).
The Patriots are likely to use either Collins or Hightower as a spy to try to prevent Wilson scrambles, just as they did against Andrew Luck in the AFC Championship Game. They are not likely to blitz very much to try to pressure Wilson. Wilson is very good against the blitz, going from 6.7 yards per play with three or four pass rushers to 7.7 yards per play with five or more.* Meanwhile, Jacksonville was the only defense in the league to send five or more pass rushers less often than the Patriots, who blitzed 20.7 percent of pass plays. The Pats actually allowed more yards per pass blitzing (6.6) than not blitzing (6.3) and they were particularly gruesome on the rare occasion that they sent a defensive back blitz (only 5.8 percent of pass plays, but they allowed an average of 9.7 yards on those 34 plays).
One more stat I couldn't fit in elsewhere above: We've written a few times about the Seahawks' tendency to start games slow and then improve in the second half. They rank 13th on offense and seventh on defense before halftime, then lead the league in both offensive and defensive DVOA after halftime. The offense, at least, has to tangle with a Patriots defense that shows the same trend. In fact, New England's trend is more specific. The Patriots this year ranked just 30th with 11.8% defensive DVOA in the first quarter of games. Only Tampa Bay and Oakland were worse defenses for the first 15 minutes. After the first quarter, the Patriots had -8.4% DVOA, making them the league's No. 7 ranked defense from the second quarter on.
WHEN THE PATRIOTS HAVE THE BALL
You may remember that our preview of the AFC Divisional matchup between Baltimore and New England pointed out that Joe Flacco seems to have his best games against the league's worst pass defenses, even after adjusting for quality of opponent. Kenneth Arthur pointed out this week in a piece over on FieldGulls.com that the opposite seems to be true for Tom Brady, at least this season. For the most part, Brady had his best games this year against the best pass defenses on the schedule. His best game by DVOA came against Chicago, ranked 29th in pass defense, but he also clobbered Buffalo (No. 1) and Cincinnati (No. 7). Not counting the Week 17 game against Buffalo, where he played only one half with a number of other starters sitting on the bench, Brady's bottom four games in unadjusted VOA all came against pass defenses ranked between No. 13 and No. 25.
This trend doesn't stick out quite as much with a larger sample, but yes, Brady really does seem to play best against his best opponents -- and his worst opponents. I took every game from the last three seasons, including the playoffs, and broke them down into four categories based on the final regular-season pass defense DVOA rank of the opponent.
|Tom Brady Pass DVOA/VOA Based on Opponent Quality, 2012-2014|
|All Games||Only Games with Gronk Active|
|Pass Ds 1-8||31.3%||15.4%||17||46.8%||31.3%||9|
|Pass Ds 9-16||14.3%||11.2%||16||6.5%||3.5%||10|
|Pass Ds 17-24||19.6%||25.6%||16||25.4%||31.5%||12|
|Pass Ds 25-32||35.1%||50.4%||5||43.5%||56.0%||4|
The numbers for the "Pass Ds 9-16" category actually go down with Gronk active because that includes the two worst games from early this season when Gronk wasn't yet fully healthy, against Miami and Kansas City. The "Pass Ds 1-8" category includes the Week 6 game from 2012 where Brady had 12.3% passing DVOA (which was -5.8% VOA before adjustment) against Seattle. And yes, there's a little bit of arbitary endpoint-picking going on here but I tried to make a big scatter plot and it just looked messy. So I went with the table.
Looking at Brady's numbers with and without Rob Gronkowski is extra important since covering the tight end has been a (relative) weakness of the Seattle defense this season. They rank only 18th in DVOA against tight ends, which could certainly present a problem when dealing with one of the best tight ends in NFL history.
Like the Patriots defense, the Patriots offense had some absurdly strong splits if you look at the first half of the season compared to the second half of the season. For example, the Patriots had the absolute worst rushing DVOA in the league through Week 9. Then came the big Jonas Gray game against the Colts followed by the return of LeGarrette Blount. As a result, the Patriots from Week 10-17 were the second best rushing team in the entire NFL, behind only the Seahawks. Blount's numbers for the entire season may not be that impressive, but that's because he had only -10.2% DVOA in Pittsburgh. He had 15.2% DVOA in New England during the regular season, with 18.3% DVOA during the playoffs.
The Seahawks' ALY numbers show them as stronger defending runs to the offense's right than the offense's left. The Pats' ALY numbers are... well, really kind of strange. The Patriots are top four in ALY in runs marked as left end, left tackle, right end, or right tackle. They are 25th in runs up the middle (listed as middle, left guard, or right guard). And those numbers actually went down, not up, when the Patriots' rushing DVOA started to improve with the return of Blount in Week 12. The Patriots went from 4.15 ALY on runs up the middle in Weeks 1-11 to 2.62 in Weeks 12-17. On other runs, they went from 4.57 ALY in Weeks 1-11 to 5.32 in Weeks 12-17.
The threat of Blount sets up the play-action pass, which the Patriots use more than you might expect. The Pats used play-action on 25 percent of passes this year, not as frequently as Seattle but still in the NFL's top ten. They got a nice bump, going from 5.9 yards per pass without play-action to 8.0 with play-action. That's 2.1 yards per play, compared to the NFL average increase of 1.3 yards per play. The Seahawks' gap between yards allowed with and without play-action was roughly around the NFL average, but of course since they're the Seahawks, that means they were better than average in both cases.
The Seattle defense also improved in the second half of the season, as players like Bobby Wagner and Kam Chancellor finally got fully healthy. If you are alarmed by the Seahawks' red zone defensive DVOA listed in the tables at the start of this preview, don't be. Our NFC Championship preview pointed out the way the Seahawks had solved this problem in the second half of the year. Here's that table again in case you missed it last week.
|Seattle Defense in Red Zone, 2014|
Seattle's improved red zone defense continued in their first two postseason games, with -36.4% DVOA in the red zone against Carolina and -7.6% DVOA against Green Bay. (The Seahawks do get docked a bit in DVOA for letting the Packers get down to the 1 twice, even though they also stopped the Packers on the 1 twice.)
The Patriots, strangely, have both improved and declined in the red zone over the course of the season. It depends if you are talking about rushing or passing. For the entire 2014 regular season, were sixth with 28.4% DVOA passing in the red zone, but 18th with -1.9% DVOA rushing. As you might expect, considering the other improvements in the New England running game since midseason, that red-zone rushing DVOA has gone up since Week 12. The Pats had -22.3% rushing DVOA in the red zone through Week 11, but have 36.6% rushing DVOA in the red zone since Week 12, including the playoffs. However, the passing rating declined in that same time period, from 87.9% passing DVOA through Week 11 to a horrid -6.9% passing DVOA in Weeks 12-20. A big part of that is Brady's two red-zone interceptions, both on passes intended for Rob Gronkowski. However, that bad passing DVOA in the red zone is all about the regular season. It breaks down into -60.7% passing DVOA in the red zone in Weeks 12-17, and then 93.8% passing DVOA in the red zone in the postseason.
The Patriots were so bad passing in the red zone in those last few weeks that it dragged down their entire passing DVOA rating. The Patriots had 46.2% passing DVOA in Weeks 1-9, third in the NFL, and then dropped to 20.8% in Weeks 10-17, eighth in the NFL. And the drop was all because of the red zone.
|New England Passing DVOA, Red Zone vs. Rest of Field, 2014|
|Weeks||Red Zone||Not Red Zone|
There's a similar trend regarding third-and-short. Overall, the Patriots ranked 30th in third/fourth-and-short DVOA this season. They were good passing in these situations early in the year, but not running, and then they were good running in the second half of the year, but not passing. Seattle's defense ranked ninth in DVOA on third-and-short. The Seahawks' real strength on defense came on first downs and third-and-long. Seattle ranked first in the NFL in DVOA in both situations, and the Patriots' offense ranked fourth in both situations.
|Seattle Cornerbacks, 2014|
Josh McDaniels has told the media that the Patriots won't deliberately avoid Richard Sherman but I doubt they'll throw much at him either. Not only is Sherman one of the best cornerbacks in the league, but the Tom Brady wasn't very good throwing to the right side of the field this season. He had 12.7% DVOA and 6.3 yards per pass to the right side, both lower than the NFL averages of 16.3% DVOA and 7.1 yards. Brady was slightly above average throwing to the left, but he really shines in two areas:
- Brady is great throwing to the middle of the field. Brady had 63.5% DVOA, much higher than the NFL average of 38.3% DVOA, although his average of 8.8 yards per pass isn't much higher than the NFL average of 8.6 yards. The Seahawks ranked eighth in the league against throws up the middle, not quite as good as they were against throws to the outside. Shockingly, despite the presence of Earl Thomas, the Seahawks were 24th against "deep middle" passes with 53.8% DVOA allowed. Once again, the spectre of Gronk rears its head. But so do the spectres of Danny Amendola and Julian Edelman. Jeremy Lane, not Richard Sherman, may be Seattle's most important cornerback in this Super Bowl.
- Brady is great at avoiding sacks. In case you are wondering how the league's "value over average" could be above average in all three directions, left, middle, and right, it's because those passes are balanced out by sacks, which don't come with a pass direction listed and are always negative. The Patriots were second in the NFL with an Adjusted Sack Rate of 4.4 percent. In the Week 5-15 period when they finally had their offensive line sorted out, before shuffling things due to injury in the final two weeks, the Patriots had a remarkably low ASR of 2.7 percent. However, while Seattle's pass rush struggled in the first half of the season, the Seahawks ranked third in the NFL from Week 10-17 with 9.6 percent ASR on defense.
The recipe for beating the Patriots in their last two Super Bowl appearances was for the New York Giants to bring intensive pressure with just their front four. There's been some discussion that perhaps Seattle is the same kind of defense that could cause Brady the same problems. But will they? With trouble getting pass pressure in the first few weeks of the season, the Seahawks started blitzing more, and they ended up blitzing on 28 percent of pass plays this year. That ranked them 11th in the NFL.* When the Seahawks do blitz, it's an intriguing matchup of strength against strength. The Seahawks allowed 6.0 net yards per pass this year with three or four pass rushers, but that dropped to 4.7 with five pass rushers and a really stingy 3.7 with six or more pass rushers. However, Tom Brady is excellent against the blitz. Brady gained 6.4 net yards per pass with three or four pass rushers, but 7.8 with five pass rushers and 7.6 with six or more.*
The best way to pressure Brady may be a strategy the Seahawks don't use much: the defensive back blitz. Brady only averaged 4.0 net yards per pass on 40 plays this year where opponents sent at least one defensive back in the pass rush. The Seahawks only sent a DB blitz on 5.5 percent of passes -- only two teams used the strategy less often -- but allowed just 4.8 net yards per pass on these plays.*
One more stat I couldn't fit in elsewhere above: You may remember that the Patriots were at the forefront of the movement towards more shotgun formations a few years ago. Well, they've gone backwards since. They used shotgun on just 47 percent of plays this year. Once upon a time that would have been close to the league lead, but this year that ranked 29th. The Pats actually like to... gasp!... put the quarterback under center. The problem with this is that shotgun formations are still better than other formations. The Pats rank sixth in the NFL in DVOA both using shotgun and not using shotgun, but the former is 23.6% DVOA and the latter is 6.4% DVOA. However, going with tighter formations and less shotgun might be the right response to the Seahawks defense, which had a better DVOA against shotgun (-17.2%, best in the NFL) than not shotgun (-14.9%, fifth in the NFL). (Note: these stats all include pistol formations in the shotgun category.)
New England's clearest advantage over the Seattle is also, of course, the least predictable part of any football game. The Patriots are more likely than the Seahawks to make steady field position gains in the kicking game, but the steady field position gains are rarely what we remember when we think of a game turning on special teams. We remember that one big play or that one big mistake. And those plays and mistakes are rarely predictable.
Plus, if we consider the small sample sizes of special teams and judge these teams based on the last two or three years rather than just one, New England's advantage is not quite as big as it looks from the tables at the start of this preview. The Seahawks, like the Patriots, ranked in the top five of our special teams ratings for both 2012 and 2013. Stephen Gostkowski had a better 2014 season than Steven Hauschka but the two kickers have been fairly equal (amd among the league's best) over a multiple-year period when it comes to field-goal percentage, field-goal distance, and the ability to boom kickoffs. The Patriots did have superior kickoff coverage this year. Not including squibs, the Patriots stopped 18 of 40 returns short of 20 yards. The Seahawks only stopped 7 of 39 returns short of 20 yards.
The Seahawks punt coverage unit is famous for never really allowing returns, but it's not as valuable as you might think if Jon Ryan isn't punting the ball far enough downfield, which is part of why the Seahawks came out as negative in both our gross punting and net punting metrics. Opposing punt returners only tried to return 17 of Ryan's punts this season. One of those was the crazy Stedman Bailey touchdown where the Rams fooled the entire Seattle punt coverage unit, but there were three other returns of 15 or more yards.
The real difference here is in the return men, one place where the Seahawks' stats prior to 2014 don't say much about their performance in this game. With Paul Richardson now out for the year, the Seahawks used Doug Baldwin on kickoff returns in the NFC Championship game, and he fumbled once while his other two returns couldn't get past the 13-yard line. Bryan Walters is better as a punt returner, in that he's average. On the other hand, the Patriots have an above-average kick returner in Danny Amendola -- their stats for the season would be better if not for a handful of mediocre kick returns by Patrick Chung early in the year -- and one of the best punt returners in NFL history in Julian Edelman. Edelman is one of just seven players in NFL history (minimum 75 punt returns) to average over 12 yards per return during his career.
We're back to the question: do we judge these teams only on the second half of the season, or on the entire season? If the Patriots run defense from the second half of the season shows up, it's going to be hard for the run-heavy Seahawks to win this game, unless the Patriots' red-zone passing problems from the second half of the season also show up. Of course, the Patriots won't have those kinds of troubles passing in the red zone if the Seattle red-zone defense from the first few weeks of the season shows up. And so on.
I will note that ESPN.com asked me to make a pick for their "expert's page," including a score. I hate picking scores of games, but I did it, and I went with Seattle, 23-20. My thought was that all things being equal, the top defenses generally shut down the top offenses more often than the top offenses take out the top defenses, as Andrew Healy showed in this Any Given Sunday article two months ago. Of course, as we showed above, that's not necessarily true where Tom Brady is concerned. My confidence in Seattle to win this game is about as strong as our playoff odds equation's confidence in New England to win this game, which is to say it's only the tiniest bit more than 50-50. But if you put a gun to my head and made me pick, I would pick the Seahawks.
DVOA (Defense-adjusted Value Over Average) breaks down each play of the season and compares it to the NFL average based on situation and opponent. You'll find it explained further here. Since DVOA measures ability to score, a negative DVOA indicates a better defense and worse offense, and a positive DVOA indicates a better offense and worse defense.
Team DVOA numbers incorporate all plays; since passing is generally more efficient than rushing, the average for passing is actually above 0% while the average for rushing is below 0%.
SPECIAL TEAMS numbers are different; they represent value in points of extra field position gained compared to NFL average. Field goal rating represents points scored compared to average kicker at same distances. All special teams numbers are adjusted by weather and altitude; the total is then translated into DVOA so it can be compared to offense and defense. Those numbers are explained here.
Each team is listed with DVOA for offense and defense, total along with rush and pass, and rank among the 32 teams in parentheses. (If the DVOA values are difficult to understand, it is easy to just look at the ranks.) We also list red zone DVOA and WEIGHTED DVOA (WEI DVOA), which is based on a formula which drops the value of games early in the season to get a better idea of how teams are playing now (explained here).