Writers of Pro Football Prospectus 2008

14 Nov 2012

FO Mailbag: AGL and the NE Secondary

Krishna Narsu: With Ras-I Dowling going on injured reserve and the Patriots' two starting safeties being injured, they've been very banged up in the secondary yet again. It seems to me to be an annual trend to see the team ravaged with injuries in the secondary. Is there any truth to this? What does Adjusted Games Lost tell us about how injured the Patriots have been in their secondary over the years? (I feel this is a problem that goes way back -- definitely as early as 2004 since that was the year they lost Ty Law and battled injuries in the secondary.)

In short, Krishna, there's plenty of truth to it. Here is a table showing the average yearly AGL for defensive backs from 2004 to 2011:

Team DB AGL Rk
CAR 3.98 1
SD 4.28 2
NYJ 4.71 3
SEA 5.73 4
TEN 5.78 5
MIA 5.98 6
PIT 6.94 7
KC 7.27 8
DEN 7.46 9
PHI 7.46 10
NO 8.22 11
ARI 8.61 12
HOU 8.68 13
OAK 9.26 14
MIN 9.58 15
TB 9.71 16
Team DB AGL Rk
JAC 9.93 17
SF 10.09 18
GB 10.31 19
ATL 10.51 20
DAL 10.67 21
NYG 11.02 22
WAS 11.34 23
CHI 12.11 24
BUF 12.21 25
CLE 13.39 26
BAL 14.81 27
STL 15.59 28
DET 15.68 29
CIN 16.70 30
NE 17.20 31
IND 22.05 32
NFL AVG 10.23 --

Remember that AGL measures the number of games lost to injury for meaningful players (i.e., not just starters) after adjusting for their status on the injury report. So, we can interpret New England's No. 31 defensive back AGL over the past eight seasons in two ways: (1) They've averaged the equivalent of losing one important defensive back for the entire season; and (2) They've suffered almost twice as many adjusted games lost as the average NFL secondary.

Of course, that's just an average. It's not like they've been decimated by injury every year -- or is it? The Patriots ranked better than 26th in defensive back AGL only twice over the past eight years. In a 2007 season that neared perfection in a myriad of ways, they ranked fourth with 1.23 AGL. Two years later, they ranked third with 2.96 AGL.

To take this a step further, I presume the underlying premise of Krishna's question to be that defensive back injuries have been a contributing factor to New England's ineffective pass defense of late. Well, I checked out the leaguewide relationship between defensive back AGL and pass defense DVOA over the eight year period, and found that the correlation equaled 0.08, which means that under one percent of the variation in the latter can be explained by the former.

That's not very surprising when we consider that (a) injuries to defensive backs don't necessarily transform a pass defense into football's version of a canned food drive, and (b) pass defense isn't just about what happens in coverage; there's also the pass rush.

In that vein, it turns out that New England's pass defense DVOA since 2004 can be summed up in the following way: As their Adjusted Sack Rate (ASR) goes, so goes their pass defense DVOA. (Feel free to check out Football Outsiders' Premium statistics database to see for yourself.) It's certainly not a one-to-one relationship, but it seems like as good a stats-based hypothesis as I can come up with on short notice.

In 2004 and 2006, high ASRs mitigated their high defensive back AGLs. In 2005, a low ASR combined with a high defensive back AGL -- the worst of both worlds -- produced the No. 31 pass defense DVOA. In 2007, a high ASR intensified the positive effect of health in their secondary.

Since then, however, middling ASRs year after year have produced no discernible pattern of pass defense performance. For instance, in 2009, a mediocre pass rush (No. 18 ASR) coupled with pristine secondary health (No. 3 AGL) resulted in an average pass defense (No. 16 DVOA). In 2010, though, a similarly mediocre pass rush (15th) resulted in a similarly mediocre pass defense (17th) despite nearly the exact opposite of pristine health in the secondary (27th).

The general interplay between AGL, ASR, and pass defense DVOA is definitely worth looking into more thoroughly when we have the down time to do so during the offseason; not to mention that we'll be able to see if New England's team-specific trend continued this year. As of this writing, they're in the middle of the ASR pack once again, and their pass defense DVOA is as well despite all the secondary injuries Krishna cited. In other words, it looks like 2010 all over again.

For now, though, there's an entire table of AGLs to chew on and discuss in the comments. Fire away.

Posted by: Danny Tuccitto on 14 Nov 2012

20 comments, Last at 19 Nov 2012, 12:18am by Insancipitory


by Guido Merkens :: Wed, 11/14/2012 - 6:50pm

Anecdotally, your hypothesis has looked strong for the Giants. Both years that they won the Super Bowl they had a ferocious pass rush make up for a beaten-up secondary.

by Independent George :: Wed, 11/14/2012 - 7:17pm

The corollary to that is that the Giants had mediocre regular season records when the secondary was injured, but made their playoff runs when the secondary got healthy.

by RickD :: Wed, 11/14/2012 - 7:08pm

People focus on the Pats' rush but ignore that often receivers are open within 2 seconds of the snap. When looking at the pass rush, you have to look at how long it takes for the Pats to generate pressure, and not merely whether they get sacks.

Back when the secondary was better, they could stay with receivers longer, and that made it easier for the pass rush to reach the QB.

by RickD :: Wed, 11/14/2012 - 7:15pm

Wondering how much of the Colts' AGL is due to Bob Sanders alone.

Answer: a lot.

From 2004-2010, Sanders played in 48 of a possible 112 games for the Colts. 64 missed games/7 years > 9 games per year, which is more than 13 teams above.

by Bobman :: Wed, 11/14/2012 - 9:58pm

Bah, Sanders just taught the other guys how to crumble. The Colts are NEVER healthy. (Look at same data for the OL & DL in particular.) I used to gripe about it until the Packers won the SB in 2010 with the only squad that had more players down that year. sigh.

by Revenge of the NURBS (not verified) :: Thu, 11/15/2012 - 1:11pm

Yeah, I didn't even need to look at the table to know the Colts would be at the bottom. Sanders is the most obvious (non)contributor. Marlin Jackson was another one. And in his last 2-3 Colts seasons, Kelvin Hayden was usually good for missing half a season. Melvin Bullitt might be another, depending on if he counts, since he played the same position as Sanders.

by theslothook :: Wed, 11/14/2012 - 10:13pm

Ah, once again, this article is a great example of the nebulous nature that is Pass Defense. For such a vitally important component of nfl football, it's so poorly understood.

I've tried tackling this issue in the past and I'll relay a few of the problems I've come up with.

1- its extremely difficult to isolate pure pass rush. ASR really isn't an effective tool and even charting hurries and hits is problematic because that can be heavily influenced by playcalling. Having just charted a colts game, I can tell you that the colts o line is positively abysmal, but when you're calling 3 step drops over and over, its hard to get much pressure.

2- Its extremely difficult trying to run regressions on Pass defense, because(and those of us who have worked with time series will understand this), you end up with cointegration issues along with other severe problems that give you really misleading results. I've tried to deal with this in the past a bit, but its been too much of a headache. I may end up revisiting it sometime in the summer.

by theslothook :: Wed, 11/14/2012 - 10:16pm

From a practical view, I find it amazing that even real nfl experts(at least the ones I listen to like charlie Casserly, Bill Polian, Jaworski, Cossell) have no idea what it takes to build a great pass defense. Do you need pass rush and what kind? is it easier to do it with blitz heavy schemes like Ryan or have superior athletes like the bears? Do you want smaller ends or penetrating tackles? And how much does coverage matter and which kinds of players? Do you want corners? Safeties? Linebackers? How many corners?

I'm convinced we have absolutely no idea how to answer any of these questions.

by Insancipitory :: Wed, 11/14/2012 - 10:42pm

There's lots of ways to do it, and it's influenced the evolution of defensive schemes, as you note. To me, the unifying theme is a good to great defensive coordinator.

When teams are putting together their defensive systems they're trying to find players undervalued by other systems while taking the shots at those rare opportunities at "Parcell's Planet Theory" people as they present themselves. So it's an economic problem, which is made progressively more difficult as teams win consistantly.

The variety and ingenuity of solutions are what I, at least, find so compelling.

by Mr Shush :: Sat, 11/17/2012 - 9:36pm

I think the 49ers may in fact have enabled a sufficiently preposterous array of talent to field a strong defense even with really quite a bad defensive co-ordinator; I doubt you'll find many Texans fans with a good word to say about Fangio.

But in general, I agree. I'd add that a really superior talent at absolutely any position on defense makes it possible to scheme to mask weaknesses in other areas. Reed, Polamalu, Revis, Urlacher, Ware, Pierre-Paul, Seymour, Tommie Harris, Wilfork - all those guys and others besides at one time or another over the last decade made it possible for their defenses to function despite major deficiencies elsewhere.

by theslothook :: Wed, 11/14/2012 - 11:52pm

I meant to say, we have absolutely no idea what the answer to these questions are. WE can, conceivably with better data and enough time, get a good picture of it.

Its tough to reconcile when its great scheme versus great talent. Take the bears, for instance. How can you tell its scheme? Their d line is deep and both starting corners are playing like hall of famers. Then you look at the 09 jets team, which to me, really only had one all pro type of player on it. Bart Scott, David Harris, Shaun Ellis, these were good to pretty good, but hardly elite. And again, their secondary was comprised of people like kerry rhodes, dwight lowery, coleman and strickland- hardly a dominant collection of players. And somehow, that 09 pass D was one of DVOAs best.

by Danish Denver-Fan :: Thu, 11/15/2012 - 8:39am

0.08 is eight percent right? Or am I missing something obvious about correlations?

by DAJ (not verified) :: Thu, 11/15/2012 - 10:58am

0.08 is the correlation coefficient, known as r. But the amount of variation explained by this variable is a different quantity - it is measured by r^2, which in this case is less than 0.01.

by asdasdadsadasd (not verified) :: Thu, 11/15/2012 - 2:34pm

Butler just won AFC Defensive player of the week ...

Something tells me, that the defensive scheme the Patriots play is one part of the problem on their back end. No matter what player you give Belichick, this defense has been awful for years (take a look: The starting defense consists almost only of players taken in round 1 or early round 2). The players or injuries just cannot be the sole problem here.

Or how do they manage to turn decent players (for example Bodden) into something like "hole in zone" covers better?

They make an awful lot of mental mistakes on the back end. Be it McCourty never being able to turn his head around, Arrington getting burned over and over. I won't even point a finger at the rookies (Tavon Wilson, Ebner, Dennard) ... I think Dennard can be a decent corner in the future, and I hope Arrington gets bumped from the starting lineup by Talib.

by Edgerat (not verified) :: Thu, 11/15/2012 - 5:38pm

Butler was also facing the Jaguars offense, next weeks game against the Pats is a much better judge of how he's turning out.

You also say it's scheme but then proceed to point out how the mental mistakes of Patriots DBs. So which is it? Scheme or just bad players?

by Alex (not verified) :: Thu, 11/15/2012 - 7:13pm

Bodden was good for one year, was on IR the next and lost motivation (or something) after that. Not dismissing your overall point, but just wanted to nitpick.

by Jimmy :: Fri, 11/16/2012 - 9:49am

Bodden was never that good. In Cleveland he had one year where his statistical production looked impressive but it was probably scheme driven. Then he stunk up Detroit for a while before having one reasonable season in NE which caused Pats fans (OK some Pats fans) to start talking about him like he was a Pro Bower they had found under a rock. You don't wind up being moved on from three of the worst defenses in football without being a poor player.

by JDL4 :: Fri, 11/16/2012 - 3:54pm

I think that NE's dependance on the quality of its pass rushers is accentuated by its defensive philosophy. NE rarely blitzes, and focuses on not getting beet deep. When things are working well, or adequately, its D gives up yards but not because it forces turnovers by making the opponent run more plays to get into the end zone. When things are not working well, opposing quarterbacks have too much time and are able to complete the big plays despite the extra men in coverage or are able to take advantage of outright mistakes in coverage by sub-par personnel.
I also suspect that the bend-don't-break strategy has become less effective because passers have become more effective at the short passing game with fewer unforced errors. Does anyone know how the league-wide interception rate has changed over the last five to ten years?

by erniecohen :: Sun, 11/18/2012 - 3:11pm

How can LBs get so little love for their role in pass defense?

by Insancipitory :: Mon, 11/19/2012 - 12:18am

Not enough really 'crispy' touchdown dances.