Bill Connelly takes a look at what we can learn from defensive box score stats and general rates of havoc.
11 Jan 2013
by Aaron Schatz
Nothing is ever guaranteed in the NFL. There is a reason for the phrase "any given Sunday." The history of the NFL postseason, in particular the recent history, is filled with surprising upsets. Nevertheless, it is difficult to remember a year when one conference championship game seemed as predetermined as the New England Patriots at the Denver Broncos, aka Brady-Manning XIV. The Patriots and Broncos finished as two of the top dozen teams in DVOA history. Their opponents, Baltimore and Houston, ranked eighth and 11th in DVOA, respectively. The Broncos had only one game all year with a single-game DVOA below 0%. The Patriots had only two such games. The Ravens and Texans each had eight.
Each of these games features a DVOA gap of over 25 percentage points. It's hard to find another year where one conference's Divisional round had two matchups where the gap was even 15 percentage points of DVOA. I went back through our numbers to look for another year in which the Divisional round featured two teams with DVOA over 25% hosting two teams with DVOA below 10%. There's only one: 1997, when the NFC Divisional round featured Green Bay (29.7%) hosting Tampa Bay (8.0%) and San Francisco (27.5%) hosting Minnesota (-4.6%). Just like with this year's AFC, these teams had played each other in December, when Green Bay beat Tampa Bay 17-6 and San Francisco beat Minnesota 28-17. A few weeks later in the playoffs, Green Bay beat Tampa Bay 21-7 and San Francisco beat Minnesota 38-22.
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. Any game charting data that appears with a asterisk appears courtesy of the ESPN Stats & Information Group and is complete through the end of the season. Game charting data is still incomplete, but represents roughly 90 percent of the season.
For a scouting perspective, make sure to also check out Andy Benoit's AFC Divisional Round Film Room.
Parts of the Baltimore-Denver preview also appeared as part of an article on ESPN Insider.
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.
It is often said that football is a game of matchups. But it may be more accurate to say that football is a game of strengths and weaknesses. Coaches examine all the matchups to figure out where they have an advantage, then try to attack the opponent's weaknesses through certain plays or tendencies. That presents a real problem for the Baltimore Ravens, because the Denver Broncos are an extremely well-rounded team without a lot of glaring weaknesses. For a playoff preview article I wrote this week for ESPN Insider, I pored through our spreadsheets looking for Denver weaknesses that the Ravens could exploit. They were hard to find, and when I did find a Denver weakness, the Ravens were often just as weak in the corresponding area.
Let's start with some of the glaring "resistable force vs. movable object" matchups. On defense, the Broncos have had a lot of trouble stopping opponents in the red zone. Though they rank fifth in our defensive ratings overall, they are 19th in the red zone, including 27th against red-zone passing. However, the Ravens offense also ranks just 19th in the red zone, and is 28th on red-zone passes.
Our Adjusted Line Yards stats, which measure how well teams block or stop runs in different directions, show that the Broncos have trouble stopping runs to the left. However, the Ravens are better running up the middle or to the right than they are running left.
With so much attention paid to Von Miller and Elvis Dumervil, not to mention Champ Bailey having a bit of a comeback season, you may not have noticed that the Broncos were equally strong against the run. This game is not going to be easy pickings for Ray Rice. The Broncos were very good at stopping long runs, leading the league with just 0.35 Open Field Yards allowed per carry (yards after the first 10) and third with just 1.03 Second Level Yards allowed per carry (yards between 5-10).
The Broncos were very strong on first down, preventing opposing running backs from putting their teams in good second and third down situations. The Broncos allowed just 3.49 yards per carry to opposing running backs on first down, the best figure in the league. However, they weren't so strong at preventing running gains on third downs, ranking just 18th in DVOA. The Ravens' running game was just 14th on third downs (as opposed to seventh overall), but that still may be the better strategy on third down with just a few yards to go, because the Broncos' defense became nigh-impenetrable with a league-leading -66.9% pass defense DVOA on third downs. The Ravens were just 21st in pass offense DVOA on third downs, primarily because they couldn't connect in situations with just 1-6 yards to go. On third-and-long, however, the Ravens ranked fourth in offensive DVOA, so that's at least a relative strength even though the Broncos were second in defensive DVOA on third-and-long.
Last week's wild card preview discussed how Joe Flacco's numbers were significantly better passing to the middle or right side of the field, rather than to the left. Here's that table again:
|Joe Flacco Passes by Direction, 2012 (includes DPI)|
On the surface, this looks like a good split for the Ravens; the Broncos were equal against shorter passes left and right, but better against deep left passes (-22.9% DVOA) than they were against deep right passes (12.9% DVOA). However, this looks like it was probably just random variation. The Broncos tend to put Bailey on a specific receiver rather than leave their cornerbacks on specific sides, so that left vs. right number doesn't reflect any particular defensive back. The Denver defensive backs were very good this year, although they certainly were helped by a pass rush that ended up leading the league in Adjusted Sack Rate. We wrote before the season that Tracy Porter would be a bad free-agent signing for the Broncos, because he had poor charting numbers in New Orleans. Well, he had poor charting numbers in Denver too -- 10.7 yards allowed per pass, 45 percent Success Rate -- but he got sick in mid-October, left the lineup, and never got his spot back because Chris Harris and Tony Carter played so well. Now Porter barely ever sees the field. Instead, you get these three defensive backs:
|Denver Cornerbacks by FO Charting Stats (collected as of 1-11-13)|
The ranks include 89 cornerbacks with at least 35 charted targets, so all three players are above average in both stats except for Carter, who is average (but not below) in yards allowed per pass. In the first Baltimore-Denver game, the Broncos generally had Champ Bailey on Torrey Smith, Chris Harris on Anquan Boldin, and Tony Carter on Tandon Doss or whoever else was in the slot, although this wasn't a hard and fast rule.
Flacco's best targets for this game are likely to be the tight ends, Ed Dickson and Dennis Pitta. The one glaring flaw in the Denver defense is an inability to cover tight ends. The Broncos ranked just 24th covering tight ends this year, with players like Jermaine Gresham and Greg Olsen burning Denver for season-high 100-yard games. In my article for ESPN, I blamed this on the aged veteran middle linebacker Keith Brooking, but a reader pointed out that the Broncos actually don't use Brooking to cover tight ends very much. I went and checked, and the reader was correct and I was mistaken. When Brooking has to cover someone, it's generally a running back, and we only have him with 14 pass targets anyway. Instead, the dismal performance against tight ends can be generally blamed on safety Mike Adams, who we have listed with 7.8 yards allowed per pass and a 42 percent Success Rate. Linebackers D.J. Williams and Danny Trevathan will also cover tight ends, although their charting numbers are better.
This was the one thing that went right for the Ravens in the first Broncos game. Dickson was injured, but Pitta caught seven of ten pass targets for 125 yards and two touchdowns. In fact, he should have caught all ten passes; the three incomplete passes were all drops. Pitta beat Adams on one of his long touchdowns, Williams on the other. However both those touchdowns came in the fourth quarter with the game essentially decided, which emphasizes the words "one thing" in the phrase "the one thing that went right for the Ravens."
Once again, here we have a place where the few Denver weaknesses generally correspond to Baltimore weaknesses. For example, the Broncos offense tends to start slowly, ranking just 12th in offensive DVOA in the first quarter. However, the Baltimore defense isn't particularly good in the first quarter either, ranking 22nd.
And as you might expect from a team that depends on Peyton Manning, the Denver offense is much better passing the ball than running the ball. They rank second in our ratings for passing, 15th for rushing. But when they really, really need to run for just a yard or two, the Broncos get it more often than not. The Broncos converted 67 percent of runs in "power" situations (i.e. third down, fourth down, or goal line with 1-2 to go), which ranks ninth in the league. Meanwhile, the Baltimore defense, which has been great against the run for over a decade, slipped badly in 2012 and ranked just 26th in our rankings. The Ravens were also awful in those short-yardage situations, allowing opponents to convert 76 percent of the time.
Even though the running game isn't a big part of what they do on offense, the Broncos are still killing opponents with play-action passing. The Broncos have gained 10.1 yards per pass with a play fake compared to just 6.4 yards per pass without a play fake.* That difference of 3.7 yards was the second-largest in the league behind Washington. And that's a huge problem for the Ravens, because play-action passes represent a huge weakness for their defense. The Ravens allowed 8.6 yards per pass on play action compared to 5.4 yards per pass without a play fake. That was the largest gap for any defense in the NFL.
In his Film Room preview of this game, Andy Benoit talked about Denver lining up using 3x1 sets in the first Baltimore game to go after cornerback Cary Williams. I definitely expect to see more of that in this game. As we noted in last week's wild card preview, former Bears special teams dynamo Corey Graham has done a pretty good job replacing Lardarius Webb as the starting cornerback on the offensive right side. The numbers we have so far give Graham a 62 percent Success Rate, 11th in the NFL among corners with at least 35 charted targets. He's allowed 5.8 yards per pass, which ranks 16th. Cary Williams, however, has 43 percent success rate (82nd) and 7.8 yards per pass (59th). The Ravens allowed 69.6% DVOA on passes marked in the play-by-play as "deep left," which was 30th in the NFL. The Broncos' offense had 65.4% DVOA on these passes, which was third in the NFL. I noted last week that Williams tended to give up a lot of short completions, but there are plenty of deep completions on his side as well. (We have a lot of these marked in our charting with the safeties in coverage; we'll have to go back and check that after the season to make sure that's accurate.)
Denver only tried two deep left passes in the first game between these teams, but one of them was the 51-yard touchdown pass where Eric Decker beat Williams, further explained in yesterday's Film Room. It's interesting to note that our numbers run counter to the conventional wisdom that Manning's deep ball improved as the year went on. His completion rate and DVOA both declined after Denver's Week 7 bye, although yards after catch improved:
|Peyton Manning on Passes 16+ Yards in the Air in 2012|
|Peyton Manning on Passes 25+ Yards in the Air in 2012|
Going back to those 3x1 sets that Andy wrote about, the Ravens did very well in our numbers against offenses lined up in what we deemed "trips" formations. Those formations are often designed to test quarters coverage, yet the Ravens had a league-low 4.2 yards allowed per pass play on 52 charted plays. However, our definition of "trips" only includes formations where three receivers are standing on one side. The Decker play, as you can see in the Film Room graphics, had three receivers on one side but one of them is a tight end in a conventional tight end crouch. Our charting would mark that as "slot right" rather than "trips right," and the Ravens allowed an above-average 6.9 yards per pass on pass plays marked as "slot left" or "slot right."
This is the place where Baltimore has a clear advantage, although it isn't like the Broncos are poor. The Ravens just happen to be very good at every aspect of special teams. They led FO's special teams ratings this season while Denver was 13th. The Broncos were 25th in punt return value, but the negative value is almost all veteran Jim Leonhard; if Trindon Holliday's ankle is healthy enough for him to go, punt returns should be fine. And the Broncos ranked just 26th in FG/XP value, as Matt Prater was just 8-of-13 on field-goal attempts over 40 yards despite kicking half the time in high altitude. Of course, in order to force Prater into a long field goal, the Ravens have to stop Peyton Manning first.
You cannot deny Brady-Manning XIV, for it is your destiny.
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.
If you look just at the ranks in those tables above, you might think that these teams are actually quite closely matched except for on special teams. But that's not the case, because the extremes in offense are always stronger than the extremes in defense. As you can see, the difference between an average offense and the best offense in the league is more than twice as big as the difference between an average defense and the third-best defense in the league. Add that to the special teams difference and home-field advantage, and you end up with New England as a heavy favorite.
You know how this works by now: The Texans offense is based around the running game and the play-action pass. The play-action pass is still a good weapon against the Patriots, who allowed 8.3 yards per play on play-action passes.* But the running game is a bit of a problem. First of all, the Houston running game actually wasn't that great this year -- just 16th in DVOA. And the Patriots had a very strong run defense, sixth in DVOA. The Patriots were particularly strong in short-yardage situations, allowing conversions just 50 percent of the time (second in the NFL).
If Arian Foster is going to get some traction against the Pats, it will likely be with runs to the outside. The Patriots ranked 30th in Adjusted Line Yards against runs around right end, and 31st against runs around left end. Foster had just 46 yards on 15 carries in the first Houston-New England game, but his two longest carries, for 15 and 9 yards, came on runs that went wide around each end.
However, if the Texans can't get the running game going, that will leave them in third-and-long situations, and as we discussed in last week's wild card preview, the Texans have been horrible in third-and-long situations all year long. Houston is 24th in offensive DVOA on third downs, and 29th when it is third-and-long (7+ yards to go).
Just like in 2011, the Patriots' secondary underwent a lot of changes during the season, but things definitely improved around midseason. First, the Patriots moved seventh-round rookie steal Alfonzo Dennard into the starting lineup in Week 7. For the season, Dennard has a 57 percent Success Rate (32nd out of 89 cornerbacks) and 7.4 yards allowed per pass (52nd). Then the Patriots acquired Aqib Talib in a trade with Tampa Bay and installed him as the left cornerback (offensive right), with Dennard becoming the permanent starter at right cornerback. Talib has surprisingly poor charting numbers since he joined the Patriots at midseason (9.3 yards allowed per pass and just a 38 percent Success Rate based on our current data), but the five games with Talib in the starting lineup were undoubtedly the best five-game stretch for the Patriots' defense all season. Talib's arrival allowed Devin McCourty to stay at safety, where he thrived, reading quarterbacks and making big hits in the open field instead of getting lost in man coverage on wide receivers. These moves also put Kyle Arrington in the slot, where he makes a better fit. Our charting numbers give Arrington 13.2 yards allowed per pass with 30 percent Success Rate through Week 10, which would be the worst numbers for any corner in the league over the course of a full season. In Weeks 11-15, Arrington allowed 4.4 yards per pass with 73 percent Success Rate.
You can see the team's improvement with each move, and then a decline when Dennard and Talib get hurt:
Andre Johnson dominates the Texans' passing game, and both Talib and Dennard will draw assignments against him. Early on in the Week 14 game, it looked like the Patriots might be specifically using Talib to cover Andre Johnson, but eventually they settled into their usual mode of leaving the cornerbacks on specific sides.
In the second half of the season, the Patriots' pass defense improved against all five of the types of receivers that we track. Nonetheless, they've had a weakness against tight ends all season long. Surprisingly, the Texans only threw three passes to Owen Daniels all game in Week 14. (He caught two, for 24 yards). He could be a bigger problem for the Patriots this week; so could the two-tight end sets that also feature Garrett Graham, who missed the first Patriots game with an injury.
One of the other changes the Patriots made around midseason was an increase in blitzes. In their first eight games, the Patriots blitzed on 16.6 percent of pass plays, and sent a big blitz (six or more) only 2.5 percent of the time*. Since Week 10, the Patriots have blitzed on 32.0 percent of pass plays, sending a big blitz 7.9 percent of the time. Matt Schaub was good against a five-man blitz this year, but struggled against the big blitz. Schaub gained 6.5 yards per play against four pass rushers, 7.7 yards per play against five, and 4.9 yards per play against six or more. However, opponents didn't send a big blitz against Houston very much -- in fact, just 4.3 percent of the time, the lowest figure in the league. Meanwhile, the Patriots' defense got stronger by sending more pass pressure: 7.1 yards per play with four pass rushers, 6.5 with five pass rushers, and 5.9 with six or more.*
You also have to ask the question whether the Patriots even need to worry about deep passes at all. A few times on Boston sports radio this week, I've heard the theory that Schaub might be hiding an injury. He was never exactly Matthew Stafford or Joe Flacco, but his ability to throw deep has completely deserted him in the last two months. Take a look at Schaub's performance on passes of 20 or more yards:
|Matt Schaub on Passes 20+ Yards in the Air in 2012|
Why did I cut the stats here at 20 yards, rather than our more common markers of "deep" (16+ yards) or "bomb" (25+ yards)? Because against Cincinnati last week, Schaub didn't even attempt a single pass that went 20 or more yards through the air.
And now, here's the real problem. In the first game between these teams, the Patriots scored touchdowns on their first three drives. They went three-and-out on their next three drives, but by that point the Texans were already scrambling to play from behind, and then the Patriots had another three touchdowns in the second half. By the way, the Patriots played that game without the best tight end in the NFL, Rob Gronkowski. Sunday will mark only the sixth game all season where the Patriots have both Aaron Hernandez and Rob Gronkowski on the field.
The Texans have the best defensive player in the game. Yet when these teams played in Week 14, J.J. Watt didn't have anywhere near the impact that you would expect, unless you are asking about how much pain Tom Brady felt on Tuesday morning. I spent the first game watching Watt on nearly every play. Watt moved all around the formation and brought non-stop pressure. He got to Brady again and again. The problem is that by the time he had gotten to Brady, Brady had usually gotten rid of the ball. The Patriots generally didn't use a double team on him, trusting Brady to dissect the defense and deliver the pass before Watt could knock him to the turf.
Last week, we noted that it was the Texans' coverage, not their pass rush, that faltered in the second half of 2012. The Texans have been having huge problems when they send just four pass rushers*:
|Texans by Number of Pass Rushers in 2012*|
|Weeks 1-7||Weeks 9-17|
|4 pass rushers||125||4.3||158||6.7|
|5 pass rushers||74||5.3||104||6.0|
|6 pass rushers||48||6.0||55||5.1|
Of course, if you big-blitz Tom Brady, he'll kill you. Brady averaged 7.8 yards per pass against six or more pass rushers. The Texans big-blitzed Brady six times in the first game. They got a sack on third down, but they also got three completions all for first downs, a total of 52 yards, and a fourth pass that would have been complete but Brandon Lloyd dropped it. The best recipe may be to go with five rushers, which is something the Texans are used to doing; they were near the top of the league using five pass rushers on 30 percent of pass plays.*
If the Texans are going to stop the Patriots' offense this week, they're going to need a better performance from veteran cornerback Johnathan Joseph. As I noted last week, Joseph struggled this year after a Week 3 injury. He didn't move with his usual smoothness, and he allowed 8.2 yards per pass after Week 3. The Texans ranked fourth in DVOA against No. 1 receivers but 28th against No. 2 receivers; however, this didn't really correlate to the specific performances of Joseph and Kareem Jackson because Joseph missed some games with the injury and, when he was healthy, the Texans would generally use him to cover whichever receiver was more of a speed threat, no matter if he was the other team's better overall receiver or not. In the first game, the Texans used Joseph exclusively on Lloyd with Jackson covering Wes Welker. Lloyd caught 7-of-9 passes for 89 yards and a touchdown, his third-highest yardage total of the year. Welker, with the much-improved Jackson covering him, caught just 3-of-9 passes, although he did gain 52 yards on those three catches.
The Texans ranked fourth in DVOA against tight ends, but as we know, Hernandez's role is much more akin to the third receiver than it is to a tight end, and as we noted last week, the Texans are seriously stretched for nickelbacks right now. Brice McCain is out for the season. Alan Ball was next man up, and he struggled. Now he's dealing with a foot injury, and that puts Brandon Harris and Roc Carmichael on the field, and they've really struggled. As a result, FO metrics show that the Texans went from the league's third-best defense against slot receivers in Weeks 1-7 to the third-worst in Weeks 9-17.
Another reason why the Patriots are more likely to use Hernandez split out rather than keep things tight with two or three players in standard tight end positions: the Texans actually were the best defense in the league against personnel packages with two tight ends, allowing just 4.2 yards per play. However, they allowed a middle-of-the-pack 6.1 yards per play against personnel packages with three or more receivers.*
Houston's run defense has actually improved despite losing inside linebacker Brian Cushing in mid-October. On the surface, this doesn't seem to be the case, because the Texans have gone from allowing 4.0 yards per carry before the bye to 4.3 yards per carry since. However, when you look at the circumstances and distance to go of these runs, it (mostly) turns out the Texans have been more successful despite allowing more yardage: 13th in run defense DVOA before the bye, and fourth since. But the Patriots always have an efficient running game, in part because defenses are so keyed to play the pass, and they were excellent at killing clock back in Week 14. Patriots running backs averaged 4.8 yards per carry after New England had taken a 21-0 lead. The Patriots will also be able to run for third downs when they need to. The Texans' defense ranks 12th on third downs, including 25th against the run. The Patriots' offense is first in the league on third downs, both rushing and passing.
Here's another colossal advantage for the Patriots, as they finished fourth in FO's special teams ratings while Houston was dead last. The biggest difference comes on kickoffs; the Patriots were the league's top team on kickoff value and were good on kick returns, while the Texans were terrible in both areas. In the Patriots' Week 14 win, Stephen Gostkowski's seven kickoffs resulted in five touchbacks, a kickoff returned to the 25 by Keshawn Martin, and one only returned to the 12.
You cannot deny Brady-Manning XIV, for it is your destiny.
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.
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).
Each team also gets a chart showing their performance this year, game-by-game, according to total DVOA. In addition to a line showing each game, another line shows the team's trend for the season, using a rolling average of the last five games. Note that even though the chart appears in the section for when each team has the ball, it represents total performance, not just offense.
28 comments, Last at 13 Jan 2013, 4:51pm by LionInAZ