DVOA by Routes: Defenses
by Scott Kacsmar
In Part I of our look at data on 12 of the most common pass routes, we focused on the production of individual receivers. In Part II, we looked at last year's qualified quarterbacks. In our final part of this study, we are turning our attention to defenses, with an emphasis on touchdowns and interceptions.
While the first two parts used receiving DVOA, which does not penalize for interceptions, the following tables will credit the defense for those picks. This is why the overall DVOA will look a bit different in this part. (Sacks and plays with no intended receiver are still absent.) Also, this is from the defense's perspective, so negative DVOA is better here.
Touchdowns and Interceptions by Route
We will focus on the same 12 routes we have been studying, but our first section is a look at touchdowns and interceptions that includes all 21 route types with at least 200 attempts last season. Pass totals include defensive pass interference penalties. The number of touchdowns differs by one from the official NFL totals due to a Kansas City score that technically went down as a rushing touchdown, when it was actually a wide receiver screen thrown to nose tackle Dontari Poe, of all people. We have 413 interceptions instead of the season's total of 415, because one pick each for Kevin Hogan and Andrew Luck did not have any route data. Those were the only two interceptions of 2016 where the NFL didn't assign an intended target. Both plays saw the quarterback get picked off in the backfield under pressure with no clear intended receiver.
In most tables in this essay, defenses will be ranked by DVOA from most negative (best) to most positive (worst). However, in this first table showing overall route data, we've ranked the stats in "offensive order." (High DVOA, high touchdown rate, and low interception rate are good and ranked closer to No. 1.) In 2016, the league's average touchdown rate was 4.30 percent, and the average interception rate was 2.26 percent (lowest in NFL history). If we view the average touchdown pass on equal ground with the average interception -- and ESPN's win probability data strongly supports that for recent seasons -- then let's consider the trade-off involved with each route. I took each route's touchdown rate and subtracted the league average of 4.30 percent. I then took each route's interception rate and subtracted it from the league average of 2.26 percent. Finally, I added the two differences together to get the trade-off in percentage points, and that is the stat by which the table is sorted, from most advantageous to least.
|2016 Pass Routes: Touchdown and Interception Rates with Trade-off|
|Check & Release||506||-1.4%||9||3,014||77.4%||1.1||6.8||16||3.16%||12||4||0.79%||7||+0.33%||12|
The corner route had the best trade-off (+9.30 percent) last season. It was originally going to be our 13th route studied, but I removed it after finding no receivers had 10-plus targets on it. The corner route is basically the opposite of a post route, breaking to the flag/pylon instead of the middle of the field. It is definitely a sound red zone strategy and had the highest touchdown rate (14.12 percent) in our study. The seam was second in touchdown rate(13.27 percent), but had an interception rate (5.31 percent) that was nearly double that of the corner route. The fade (10.88 percent) was the third route to have a touchdown rate over 10.0 percent, and also had a lower interception rate (2.30 percent) than the corner (2.78 percent) and seam. These are definitely red zone strategies, which would boost the touchdown rate a bit, but that is something to be looked at another time.
The comeback (0.52 percent), swing (1.26 percent), and curl (1.41 percent) had the lowest touchdown rates, which makes sense with the comeback and curl requiring the receiver to turn his back to the defender. Swing passes, often to backs, are generally safe plays, with just one interception on 397 attempts last year.
Speaking of safe, "chip" routes had the lowest interception rate (0.24 percent). This is when a running back or tight end chips a pass-rusher before making himself available for a short pass. Only 24 of the 424 chips last season went to a wide receiver, so this is just another safe type of short throw. We also have our first look at running back screens last season, which had a much better DVOA (-1.4%) than the more common wide receiver screens (-21.1%). The touchdown rates were also nearly identical
The post route had the highest interception rate (5.63 percent), but still produced the second-best DVOA (30.8%) against defenses even with those picks factored in, as well as the fourth-best trade-off (+2.11%). Go routes were picked off at the second-highest rate (5.46 percent), but a lot of those come deep down the field. The average go route interception was returned 10.2 yards, below the league average of all interception returns (12.8 yards).
The worst trade-off belongs to the dig route, which only ranked 18th in touchdown rate (1.71 percent) and 17th in interception rate (3.19 percent). It still has one of the better DVOAs (5.5%) against defenses, but when it comes to throwing more touchdowns and limiting interceptions, there are better options.
Defenses vs. the 12 Most Common Routes
We did not include a column for touchdowns or interceptions in each route type due to the numbers being fairly small for everyone. There were only 29 instances where a team allowed at least four touchdowns on a certain route type. There were only six instances of at least four interceptions on a certain route type. There were 10 interceptions where "No Video" was listed as the play, and half of those were by the Baltimore defense. We'll have to check into what the issue was there.
The curl route was the league's most common throw last year, with 2,615 attempts according to SIS.
Well, we found something the Patriots were last in: DVOA against curl routes (25.4%). Perhaps this is by design with Bill Belichick, master of the bend-but-don't-break defense, willing to keep the play in front of his corners, and give up the shorter completions before tightening up in the red zone. Seattle, 31st in DVOA, could be viewed in a similar fashion, though a strong secondary like Denver still finished No. 2 in DVOA (-56.3%). Denver was challenged the least this way with 53 curls. The Panthers (116) and Falcons (113) faced the most curls, which New Orleans' Michael Thomas should enjoy, since he caught 26-of-28 curls last season. The Broncos and Vikings were the only defenses to allow a completion rate below 60 percent on curl routes. Dallas was last, allowing a completion percentage of 85.6 percent.
The 49ers, with a no-name secondary, are a surprise team to see on top in DVOA. That's largely because they had six interceptions on curls, the most interceptions by any defense against any route type in 2016. Denver (four) was the only other team with more than three picks on curls.
Philadelphia allowed the lowest YAC (1.6) on curls, while the Raiders (4.8) allowed the most YAC.
The out, or quick out, had 2,075 attempts last season.
Praise for the San Francisco defense in 2016 is rare, so let's quickly do it again. The 49ers allowed a league-low 52.1 completion rate against out routes. Unfortunately, they only had one interception here instead of six, and allowed a league-worst six touchdowns on out routes (no other defense had more than four).
The Dallas defense allowed a league-high 81.4 completion rate on out routes. Between this and the league-high 85.6 percent rate on curls, it makes a lot more sense why Dallas was flirting with the NFL record for worst completion percentage allowed in a season. This was even with getting two games against Eli Manning, the worst 2016 quarterback on the two most commonly used routes. Dallas settled down at 67.1 percent, and it was actually the 2016 Lions who set the record at 72.7 percent. Detroit finished 30th in completion percentage against both curls (80.0 percent) and out routes (74.0 percent). The Lions (4.3) were also the only defense to allow more than 4.0 YAC per completion on out routes. The Bears (1.5) allowed the least YAC on out routes.
The Chargers (-41.7%) and Buccaneers (-40.4%) were the best in DVOA against out routes, while the Browns (32.2%) and Saints (30.4%) were the only defenses above 30.0%. There is a lot of buzz about the potential on defense for the Chargers and Buccaneers this season, though players must stay healthy.
Arizona faced a league-high 92 out routes, barely edging out NFC West foes the Rams (90). However, Seattle faced a league-low 38 out routes.
The third-most common route was the dig route at 1,348 attempts. You can also think of this play as a "square-in" route. No defense faced more of these plays in 2016 than Oakland (62), but was that wise of their opponents?
Wow, Dallas. That's three routes where the Cowboys ranked 31st or 32nd in completion rate, and here they were 32nd in DVOA (a very bad 78.0%). Fortunately for Dallas, only Jacksonville (27) faced fewer dig routes than Dallas' 29. Brandon Carr was the main culprit on the dig route, allowing five completions on six targets for 94 yards and a touchdown. The only incompletion was dropped by Green Bay's Davante Adams. Carr is no longer with the Cowboys, but Byron Jones is a starter. He was beat on all four of his dig routes for 67 yards.
Seattle allowed the highest completion rate (76.9 percent) on dig routes. That may not be as simple of a fix as "get Earl Thomas back healthy." Thomas allowed completions on both dig routes for which he was the primary defender, and DeShawn Shead allowed quarterbacks to go 5-for-5 on digs against him.
Denver allowed a league-low 41.5 completion rate against digs, but finished second in DVOA to the Giants (-72.9%). The Giants faced the deepest digs (13.1) in the league, while the Broncos (8.2) saw the shortest. New York grabbed a league-high five interceptions on dig routes. Only Miami (four) had more than three. Janoris Jenkins was especially valuable to this stat, as he had five stops on six dig routes with two interceptions.
Miami was the only defense to face 60 slants last season. There were 1,259 attempts in total, but that's still less than half of the number of curls (2,615). With an interception rate of only 2.46 percent, the slant is a high-percentage, effective play to utilize.
Didn't really expect to see the 2016 Bears finish on top in DVOA anywhere, but they were No. 1 against slants, and also faced a league-low 23 of them. They were one of five defenses to hold passers under 60.0 percent on slants. The Texans and Jaguars allowed just a 55.6 percent completion rate on slants.
Dallas finally ranked a more respectable 18th in completion rate, but still 27th in DVOA after allowing a league-high five slant touchdowns. The Lions allowed 86.1 percent of slants to be completed, the worst rate in the league, but were 31st in DVOA rather than last.
Pittsburgh ranked last in DVOA (50.8%), and yes, that 95-yard slant touchdown by Mike Wallace had a lot to do with it. Rookie cornerback Artie Burns was in coverage there, and he faced 11 slants last year while no other Pittsburgh defender saw more than five. That was the one play that really got away from him, as the other 10 plays never gained more than 12 yards. But even with that play, Pittsburgh did not rank last in YAC allowed on slants. That would be Baltimore (9.3), as New England back James White notably got loose for 53 YAC on one play, and Odell Beckham Jr. picked up 60 YAC on a game-winning touchdown on fourth-and-1.
The drag route produces the most YAC (5.9) on average among the five most common routes in the NFL. The only defenses to face 50-plus drags last season were the Giants (56), Falcons (51), and Patriots (51). The Rams and Texans only faced 25 drags each, the lowest total.
We mentioned that the Texans and Jaguars allowed the lowest completion rate against slants, but here they are two of three defenses (the Ravens are the other) to allow higher than 80.0 percent completion rates on drag routes.
[ad placeholder 3]
Cleveland was basically the NFL average in completion rate allowed (70.3 percent) on drags, but last in DVOA (47.1%) after allowing a league-worst six touchdowns. Only Miami (four) allowed more than three scores on drag routes. Quarterbacks had a success rate of 9-for-10 with four touchdowns when targeting linebacker Christian Kirksey and cornerback Joe Haden on drag routes.
The YAC range was at its widest here, ranging from the Rams (3.1) to the Bills (10.2). Maybe those teams should make another trade or something. Buffalo, 31st in DVOA, was pretty far out there, as the next closest team in YAC was Arizona (8.4). Linebacker Zach Brown faced nine drag routes and stopped four of those plays from being successful, but Miami's DeVante Parker got loose for 52 YAC and a touchdown one time.
The Broncos, Bengals, and Titans were the only defenses to intercept two drag routes. While we are used to seeing Denver be great here, we finally see Dallas stepping up with a No. 2 ranking in DVOA (-53.9%). Byron Jones was 6-for-8 at defending drag routes to lead the team.
We are now moving into routes that had fewer than 1,000 attempts last season. There were 934 of these go/fly routes last season.
The Titans were middle-of-the-road on defense against the deep ball, but offenses still chucked them up a league-high 53 times. The next closest team was New England at 45 attempts, and also with middling results (16th in DVOA, 12th in completion rate). The Patriots were flagged a league-high six times for pass interference on go routes. However, the Titans and Patriots were two of seven defenses with three interceptions on go routes. No defense had four picks on go routes.
Cincinnati had the best DVOA (-88.9%) against go routes, but also faced a league-low 11 of them. Dre Kirkpatrick was the target on seven of those 11 plays , but not on the lone completion: a 55-yard gain from Trevor Siemian to Demaryius Thomas.
The Saints and Browns each allowed six touchdowns on go routes, more than double any other defense in the league, but neither team was last in DVOA. That distinction belongs to the Rams. Los Angeles' 16 go routes were the second-fewest, but this was the only defense to allow a completion rate of 50 percent, giving the Rams the worst DVOA (101.8%) in the process.
Wide Receiver Screen
When you have a throw that is completed 89.3 percent of the time, the best defense against it is to tackle the receiver immediately. The correlation between DVOA and YAC on wide receiver screens is 0.47. Only the drag route (0.56) had higher correlation in 2016.
|NFL WR Screen||-21.1%||-||797||4,567||89.3%||-1.7||8.1|
The Giants and Rams each faced a league-high 40 screens, while the Patriots faced a league-low eight screens. The next closest defenses were Denver (15) and Seattle (16), two secondaries with great reputations. Perhaps with New England this is a case of playing so often with the lead that opponents did not have time to be cute. Then again, the Falcons had the second-highest average lead and still faced 33 screens last season. They also struggled against those plays, ranking 31st in DVOA.
The Titans allowed a league-low 70.6 completion rate on wide receiver screens. That was helped by forcing Philip Rivers and Matthew Stafford to each throw a screen away. The Bengals (18) and Jaguars (23) had every wide receiver screen attempted against them completed, but tackled well to finish in decent standing in DVOA. Only five defenses allowed multiple touchdowns on wide receiver screens, but no one allowed three such scores. The Cardinals and Panthers were the only defenses to intercept a wide receiver screen, but Arizona also allowed the most YAC (11.4) on these plays.
The Chiefs (-57.7%), Browns (-57.6%), and Steelers (-57.6%) were incredibly close together for the DVOA lead against wide receiver screens. The Browns and Steelers were both able to bat down a pair of screens, while the only incompletion against Kansas City was thrown away. Pittsburgh allowed the least YAC (4.7) of any defense.
The post route is another longer throw in the game (average: 21.6 yards). No route produced more touchdown passes (73) in 2016.
Chicago allowed seven touchdowns on post routes, the most touchdowns allowed on any route type by a defense in 2016. Miami (six) was the only other defense to allow more than five scores. The Bears also allowed the highest completion rate (69.0 percent) on post routes. Part of what saves Chicago from having the worst DVOA here was two interceptions. The Jets (one) and Browns (zero) combined for one interception against post routes. Cleveland left a big hole open in its zone defense six times for opponents, leading to five grabs for 138 yards and one dropped post route.
Washington finished No. 1 in DVOA (-47.6%). Josh Norman did not have the impact he hoped for last season, but he only allowed a 13-yard completion to Dez Bryant on five post targets. He also came away with an interception while guarding Terrelle Pryor, a new teammate for Norman in 2017.
Cincinnati finished second in DVOA (-45.9%) and first in completion rate against (26.3 percent). Kirkpatrick was impressive again, only getting flagged for a 14-yard pass interference penalty and forcing five incompletions.
Comebacks have the lowest average YAC (1.3) of any route type with at least 50 attempts.
Arizona allowed an 83.3 percent completion rate on comebacks, but only faced a league-low six of these routes. Comebacks were used most against the Titans (35), which makes sense given the cushions on the outside that Dick LeBeau defenses are notorious for allowing. Perrish Cox was picked on with 11 of these plays, only producing a 27 percent success rate for the Titans.
The Buccaneers allowed just a 40.0 percent completion rate on comebacks, best in the league. Vernon Hargreaves and Brent Grimes only allowed two successful plays on seven comeback targets. Tampa Bay was only outdone in DVOA by Buffalo, one of five defenses with an interception on a comeback.
These are not broken plays like aborted snaps, but plays where the quarterback scrambles and the receivers break their original routes. This is backyard/sandlot football in number format.
|NFL Broken Plays||-34.4%||-||567||3,190||41.1%||11.1||3.8|
The Colts faced a league-low seven of these plays last season, but I guess that's what happens when you are the worst defense at generating pressure. They also allowed five completions for the worst completion rate (71.4 percent) and the second-worst DVOA (43.2%).
[ad placeholder 4]
The Patriots (29) actually faced the most broken plays thanks to five each from Tyrod Taylor and Colin Kaepernick. They only allowed six first downs on those 29 plays, good for a top 10 finish in DVOA.
The Jaguars, Chiefs, Rams, and Eagles allowed a league-high three touchdowns on broken plays. The Browns, Cowboys, Texans, Dolphins, and Giants tied for the league lead with two interceptions on broken plays.
We rarely get to say anything positive about the 1-15 Browns here, but Cleveland finished No. 1 in DVOA against broken plays (-189.6%). That's what happens when you face 11 plays and come away with two interceptions while only allowing one first down.
Pittsburgh (55.6%) had the worst DVOA against broken plays, allowing a first down on 8-of-16 plays, including one infamous touchdown from Carson Wentz to Darren Sproles for 73 yards in Week 3.
After the wide receiver screen, this seems to be the least favorite strategy of many fans, but the trade-off was surprisingly strong (+6.53 percent) when you're looking at balancing touchdowns to interceptions. Still, the completion rate leaves a lot to be desired, and it has the fourth-worst DVOA for offenses (-15.1%).
If 29 defenses can force an incompletion more often than they allow a catch on fade routes, that should tell you this isn't a very high-percentage play. Even the Dallas defense owned the fade, allowing one completion on 20 attempts (5.0 percent) without even the benefit of one drop. Most of that work was done by the now departed duo of Brandon Carr and Morris Claiborne, who combined to allow one completion on 14 fade routes.
Green Bay faced a league-high 29 fade routes. The Packers tied the Rams with a league-high five touchdowns allowed on fades, but also tied the Chiefs with a league-high two interceptions on fades. Ladarius Gunter and Micah Hyde stopped 8-of-12 fades, but Damarious Randall and Quinten Rollins allowed 6-of-9 fades to be successful.
I guess 2016 wasn't the year of the fade for the Panthers, seeing as how the offense struggled mightily with these throws in addition to the defense. Carolina barely edged out Tennessee for the worst DVOA and highest completion percentage against fades, though the Panthers only faced six fades. The Titans faced 16, and Cox and LeShaun Sims allowed 7-of-10 fades to be successful.
For our final route, we look at the effective seam route. The Patriots with Tom Brady were the top offense for seam routes, but how did the defense fare?
Well, it was only 13 plays, but the Patriots actually ranked last in DVOA (117.9%) against seam passes. New England was one of seven defenses to allow at least three touchdown passes on seams, but only Miami allowed four. Carolina and Pittsburgh led the league with three seam interceptions.
The Chargers were the only defense to face more than 20 seam passes. The Eagles faced just three all season. Seattle saw seven passes sail down the seam, allowing one 12-yard completion for the best DVOA (-85.1%) here.
Like we did for the quarterbacks, here is the average rank for each defense in DVOA over the 12 routes.
|Defense: Average DVOA Rank|
You could say that the Broncos and Lions were in their own classes, but the Falcons and Buccaneers surprise a little here as well. This study obviously did not include sacks, but Tampa Bay finished as the No. 6 pass defense in DVOA last year. Here, the Buccaneers finished in the top seven in six routes, while the Falcons were 23rd or worse in eight of the 12 routes. It tends to get lost in the 28-3 Super Bowl comeback because of the plays Atlanta didn't run on offense late, but that was not a good defense last season. The return of Desmond Trufant and development of many young players will be keys to getting the Falcons back to the big game.
Special thanks again to Sports Info Solutions for this route data, and we look forward to having more of it in the future. With one year of data, there is plenty of uncertainty on just how much correlation there will be year to year, but I have to imagine we'll see plays such as post routes as consistently beneficial and other plays such as curl routes falling into below-average territory for offensive efficiency.
This is data that can help us build better statistical profiles of playing styles, especially strengths and weaknesses for quarterbacks and receivers. We also might be able to build a profile for the offensive scheme of offensive coordinators, and how that might translate into a player's role when he joins a certain team. As for implementing this data into season projections for a wide receiver's DVOA, I don't think we are anywhere close to that step. However, I certainly think there has been an evolutionary process going on with how we use data to analyze a play in the passing game.
- Phase 1 (traditional stats): There was a 6-yard completion.
- Phase 2 (early FO era): There was a 6-yard completion on third-and-5 at the 50-yard line in a tied game with 10:00 left in the third quarter against the No. 12 pass defense.
- Phase 3 (charting era): This 6-yard completion was a slant thrown 4 yards beyond the line of scrimmage with 2 yards after the catch. [Cornerback] was the primary defender in coverage, the defense rushed four, and the quarterback was not under pressure.
- Phase 4 ("Next Gen Stats"): Advanced video analysis and GPS-like tracking of players can tell us how long the quarterback held the ball, how fast the receiver was running, and how much separation he gained, or how tight the coverage was from the defensive perspective.
We are pretty deep into Phase 3 now, but there is still always improvement and more to be done with charting. Phase 4 is perhaps the future of advanced NFL analysis, but that future also hinges on just how much of the data becomes available to people like us. If teams aren't going to let this data be released for the tens of thousands of plays each season, then we'll likely be limited to small offerings of things such as "fastest touchdowns of the year" and some random passing charts.
Sorting out this data for what is important, what is predictive, and what is useful will take years to achieve. At the very least, we have come a long way from using "yards gained" as a way to judge a play, and it is a good thing for us all that people continue to want to quantify more and more of what happens on each play.