The Top-5 Defenses Normalized

[PF_SU]

Wondering how the top-5 scoring defenses are doing on average I collected all data grouped into halves and then normalized them by cutting off the two halves that exceeded one standard deviation of the average points allowed in both directions. So for example, for the Patriots the first halves of the Texans and first Bills game was removed (best and worst defensive first half) and for the second halves again the Texans and the second half of the week 2 matchup against Miami was removed.

The result was this:

A couple of observations:

• The Patriots pass defense is significantly better in the second half. In terms of yards per attempt only the Eagles defense performs better in the second half. Part of it is surely the fact that the Patriots have entered the second half leading in all but one game so they are trading the rush defense for better pass coverage.

• The run defense of the Patriots is pretty good. Non-QBs have only had five runs of 15+ yards for 120 yards total (three to Gillislee alone) with one of those runs coming late in the fourth quarter by Lamar Miller with the score at 27-0. Last season the defense allowed 17 such runs for 381 yards.

• Pass rush and sack numbers are the defense's biggest problem, which is why the Patriots rank last in net yards per attempt. The Patriots coverage, despite the insecurity at the CB2 position, is good, in pure yards per attempt they are in the same region as Seattle and better than the Eagles. The lack of a quick pass rush makes them vulnerable to a dink-and-dunk style, but downfield coverage is very good. The Patriots have allowed seven pass playes of 30+ yards, the same number as Denver and fewer than Seattle (eight) and Philly (nine). Only the Vikings have allowed fewer with five. All three NFC teams though have already had their bye and thus one game less.

• With the improvement the last few weeks the normalized Patriots' 3rd down defense is actually quite respectable at 36.7%, right in the range of the other top defenses.

• Denver's defense is pretty insane against the pass. No other team has that combination of shutdown corners and a great pass rush. But unlike last year they are vulnerable to the run. Denver's allowed nine runs of 15+ yards to non-QBs for a total of 228 yards. That's already more than last year when they had eight such runs for 190 yards.

• The Eagles defense and their league leading DVOA is quite inconsistent. Of that bunch they have by far the worst first half defense while their second half defense is absolutely stifling. That inconsistency is especially present in the first half with the standard deviation of the points allowed being 7.72. The Eagles have given up 21 first half points to both the Redskins and Lions.

• The Seahawks probably have the best-balanced defense. While their pass defense isn't as overwhelming as it used to be they have a good pass rush and are a very good run D (only three 15+ yard runs allowed). Teams have a hard time moving the ball against them and when they make it into the red zone it gets even harder for them to pick up TDs.

• The Seahawks and Patriots are both at the top of first half points per drive while being on the bottom for second half points per drive. Probably has something to do with them being frontrunners. The Patriots are outscoring their opponents 74-17 in the first quarter and the Seahawks 31-6 (the halftime score is 115-57 for the Patriots and 78-41 for the Seahawks). The only quarter the Patriots lose on aggregate is the fourth with a 36-41 scoreline (the Seahawks lose both quarters in the second half, 12-24 and 38-41).

• As good as the Patriots (and Seahawks) are in the first half, the same counts for the Broncos and Eagles in the second. Denver is 91-24 in the fourth quarter (they are 16-54 in the first). The Eagles meanwhile are 67-9 in the third.

 

[Nikolai]

Good, solid analysis. Great work.

 

[Patjew]

Jeez man!

 

[RobertWeathers]

Wondering how the top-5 scoring defenses are doing on average I collected all data grouped into halves and then normalized them by cutting off the two halves that exceeded one standard deviation of the average points allowed in both directions. So for example, for the Patriots the first halves of the Texans and first Bills game was removed (best and worst defensive first half) and for the second halves again the Texans and the second half of the week 2 matchup against Miami was removed.

The result was this:

A couple of observations:

• The Patriots pass defense is significantly better in the second half. In terms of yards per attempt only the Eagles defense performs better in the second half. Part of it is surely the fact that the Patriots have entered the second half leading in all but one game so they are trading the rush defense for better pass coverage.

• The run defense of the Patriots is pretty good. Non-QBs have only had five runs of 15+ yards for 120 yards total (three to Gillislee alone) with one of those runs coming late in the fourth quarter by Lamar Miller with the score at 27-0. Last season the defense allowed 17 such runs for 381 yards.

• Pass rush and sack numbers are the defense's biggest problem, which is why the Patriots rank last in net yards per attempt. The Patriots coverage, despite the insecurity at the CB2 position, is good, in pure yards per attempt they are in the same region as Seattle and better than the Eagles. The lack of a quick pass rush makes them vulnerable to a dink-and-dunk style, but downfield coverage is very good. The Patriots have allowed seven pass playes of 30+ yards, the same number as Denver and fewer than Seattle (eight) and Philly (nine). Only the Vikings have allowed fewer with five. All three NFC teams though have already had their bye and thus one game less.

• With the improvement the last few weeks the normalized Patriots' 3rd down defense is actually quite respectable at 36.7%, right in the range of the other top defenses.

• Denver's defense is pretty insane against the pass. No other team has that combination of shutdown corners and a great pass rush. But unlike last year they are vulnerable to the run. Denver's allowed nine runs of 15+ yards to non-QBs for a total of 228 yards. That's already more than last year when they had eight such runs for 190 yards.

• The Eagles defense and their league leading DVOA is quite inconsistent. Of that bunch they have by far the worst first half defense while their second half defense is absolutely stifling. That inconsistency is especially present in the first half with the standard deviation of the points allowed being 7.72. The Eagles have given up 21 first half points to both the Redskins and Lions.

• The Seahawks probably have the best-balanced defense. While their pass defense isn't as overwhelming as it used to be they have a good pass rush and are a very good run D (only three 15+ yard runs allowed). Teams have a hard time moving the ball against them and when they make it into the red zone it gets even harder for them to pick up TDs.

• The Seahawks and Patriots are both at the top of first half points per drive while being on the bottom for second half points per drive. Probably has something to do with them being frontrunners. The Patriots are outscoring their opponents 74-17 in the first quarter and the Seahawks 31-6 (the halftime score is 115-57 for the Patriots and 78-41 for the Seahawks). The only quarter the Patriots lose on aggregate is the fourth with a 36-41 scoreline (the Seahawks lose both quarters in the second half, 12-24 and 38-41).

• As good as the Patriots (and Seahawks) are in the first half, the same counts for the Broncos and Eagles in the second. Denver is 91-24 in the fourth quarter (they are 16-54 in the first). The Eagles meanwhile are 67-9 in the third.

The defense still sucks.

 

[SEA_Pat]

Nice work, very helpful to put things in perspective.

I suppose I should now start talking about the availability of Dion Lewis here...

 

[bbobbo]

Wondering how the top-5 scoring defenses are doing on average I collected all data grouped into halves and then normalized them by cutting off the two halves that exceeded one standard deviation of the average points allowed in both directions. So for example, for the Patriots the first halves of the Texans and first Bills game was removed (best and worst defensive first half) and for the second halves again the Texans and the second half of the week 2 matchup against Miami was removed.

The result was this:

A couple of observations:

• The Patriots pass defense is significantly better in the second half. In terms of yards per attempt only the Eagles defense performs better in the second half. Part of it is surely the fact that the Patriots have entered the second half leading in all but one game so they are trading the rush defense for better pass coverage.

• The run defense of the Patriots is pretty good. Non-QBs have only had five runs of 15+ yards for 120 yards total (three to Gillislee alone) with one of those runs coming late in the fourth quarter by Lamar Miller with the score at 27-0. Last season the defense allowed 17 such runs for 381 yards.

• Pass rush and sack numbers are the defense's biggest problem, which is why the Patriots rank last in net yards per attempt. The Patriots coverage, despite the insecurity at the CB2 position, is good, in pure yards per attempt they are in the same region as Seattle and better than the Eagles. The lack of a quick pass rush makes them vulnerable to a dink-and-dunk style, but downfield coverage is very good. The Patriots have allowed seven pass playes of 30+ yards, the same number as Denver and fewer than Seattle (eight) and Philly (nine). Only the Vikings have allowed fewer with five. All three NFC teams though have already had their bye and thus one game less.

• With the improvement the last few weeks the normalized Patriots' 3rd down defense is actually quite respectable at 36.7%, right in the range of the other top defenses.

• Denver's defense is pretty insane against the pass. No other team has that combination of shutdown corners and a great pass rush. But unlike last year they are vulnerable to the run. Denver's allowed nine runs of 15+ yards to non-QBs for a total of 228 yards. That's already more than last year when they had eight such runs for 190 yards.

• The Eagles defense and their league leading DVOA is quite inconsistent. Of that bunch they have by far the worst first half defense while their second half defense is absolutely stifling. That inconsistency is especially present in the first half with the standard deviation of the points allowed being 7.72. The Eagles have given up 21 first half points to both the Redskins and Lions.

• The Seahawks probably have the best-balanced defense. While their pass defense isn't as overwhelming as it used to be they have a good pass rush and are a very good run D (only three 15+ yard runs allowed). Teams have a hard time moving the ball against them and when they make it into the red zone it gets even harder for them to pick up TDs.

• The Seahawks and Patriots are both at the top of first half points per drive while being on the bottom for second half points per drive. Probably has something to do with them being frontrunners. The Patriots are outscoring their opponents 74-17 in the first quarter and the Seahawks 31-6 (the halftime score is 115-57 for the Patriots and 78-41 for the Seahawks). The only quarter the Patriots lose on aggregate is the fourth with a 36-41 scoreline (the Seahawks lose both quarters in the second half, 12-24 and 38-41).

• As good as the Patriots (and Seahawks) are in the first half, the same counts for the Broncos and Eagles in the second. Denver is 91-24 in the fourth quarter (they are 16-54 in the first). The Eagles meanwhile are 67-9 in the third.

very nice.

i posted this in the "excellent reiss piece" thread a little while before you created this thread:

Excellent Reiss piece about the trade

i'll copy it here since this seems to be a more appropriate place for it.

----------------------------------------------------------------------------------------------------------------------
some context on the patriots #3 defensive ranking on points:

• the only teams the patriots have faced that are in the top half of the league in scoring offense are buffalo twice (#8) and a steelers team without roethlisberger (#13)
• the dolphins average 13.7 points on the road but scored 24 against the patriots
• the texans only score 7.3 points on the road, so the shutout against them seems less impressive
• the patriots were playing against the browns' 3rd and 4th (and 5th, at times) string QB
  
• the bengals only score 17.5 points on the road, so 17 points scored against the patriots is average

Through Week 8

(home/away average is the opposing team's scoring average where the patriots played them)
----------------------------------------------------------------------------------------------------------------------

any chance you could perform the same analysis for the other top 5 defenses?

 

[PF_SU]

the texans only score 7.3 points on the road, so the shutout against them seems less impressive

The Texans aren't very good on offense (Osweiler, heh), but you failed to mention who they played on the road (all losses): @Patriots, @Vikings, @Broncos... ouch. @Patriots they had 263 net yards and 0 points (24-41, 196 for Osweiler). @Vikings they had 263 net yards and 13 points (19-42, 184 for Osweiler). And @Broncos they had 272 net yards and 9 points (18-34, 107 yards for Osweiler, deducted the couple passes he made at the end in my analysis above since that was run-out-the-clock mode, like the couple runs by Alfred Blue against New England).

any chance you could perform the same analysis for the other top 5 defenses?

Let me come back to that later.

 

[ALP]

if you wouldn't mind, would you do analysis after this weekend again when the Pats have also had a bye?

that way all 5 teams would have had one and the sample sizes would be similar

seems like only thing holding the NE-SEA game from being a very good one next week might be the inept SEA offense

 

[PF_SU]

any chance you could perform the same analysis for the other top 5 defenses?

Well, I thought about doing it, but in the end decided against it. The problem here is that not every difference is the same. For example the Panthers average 35.5 points in their two road games not including Denver. But these two lines came against the Falcons and Saints, two of the worst defenses in the NFL. Meanwhile the Texans average 11 points in their two road games not including Foxborough, but as mentioned above, these were in Minnesota and Denver.

So if you want a cross-comparison factoring in the strength of the opposition I'm refering you to Pro-Football-Reference's SRS. Their SRS is basically what you are looking for. Seattle's #1 in defensive SRS at 8.4, Denver #2 at 8.3, Minnesota #3 at 7.4, New England #4 at 6.9 and Philadelphia #5 at 6.2.

 

[Patjew]

Well, I thought about doing it, but in the end decided against it. The problem here is that not every difference is the same. For example the Panthers average 35.5 points in their two road games not including Denver. But these two lines came against the Falcons and Saints, two of the worst defenses in the NFL. Meanwhile the Texans average 11 points in their two road games not including Foxborough, but as mentioned above, these were in Minnesota and Denver.

So if you want a cross-comparison factoring in the strength of the opposition I'm refering you to Pro-Football-Reference's SRS. Their SRS is basically what you are looking for. Seattle's #1 in defensive SRS at 8.4, Denver #2 at 8.3, Minnesota #3 at 7.4, New England #4 at 6.9 and Philadelphia #5 at 6.2.

I just want to commend you on an incredible sentence:

"The problem here is that not every difference is the same."

Love it.

 

[aluminum seats]

Very Sons of Sam Horn here...

 

[oldrover]

Wondering how the top-5 scoring defenses are doing on average I collected all data grouped into halves and normalized...

...aaaaaaand that's where you lost me. LOL (nice work)

 

[luuked]

Pretty cool analysis..

Not sure I would call it normalization but more "outlier" removal. It shows some interesting trends, however the one of the core issue with all of that is that every team played a different set of teams so the stats you generated dont really represent an apple to apple comparison.

Normalization is exactly what the data would need to be more comparable, but that's pretty complex business (what do you normalize against ?) and I suppose that the DVOA metric is pretty good already.

In any case, I didn't wanna critizise your efforts because I am pretty thankful you invested some time into that.

 

[PF_SU]

Not sure I would call it normalization but more "outlier" removal. It shows some interesting trends, however the one of the core issue with all of that is that every team played a different set of teams so the stats you generated dont really represent an apple to apple comparison.

Normalization is exactly what the data would need to be more comparable, but that's pretty complex business (what do you normalize against ?) and I suppose that the DVOA metric is pretty good already.

Aye, it's just an overview of what you can expect of a defense when they perform at their "usual" level. Also did it to see how the game situation affects the defensive play of different teams. Analysis that use a full game's data sometimes have a hard time distinguishing that.

In regard to DVOA, it's a well thought-out idea, but it's yardage-based. So I'm prefering to use it together with Defensive SRS, which is also designed to ensure comparability of different sets of data, which is based on points relative to average. So take the 2006 Patriots for example, #2 in total points allowed, #3 in points per drive. Their DSRS is #2 behind Baltimore. Meanwhile their DVOA is #7 because they finished #13 in yards per drive. They rank behind the 2006 Panthers for example who had a significantly worse points per drive figure (#11) but were #7 in yards per drive.

The Patriots DVOA is currently #19, dragged down by the horrific second half of week 2 against Miami (since week 3 the Patriots are #11 in DVOA) while their DSRS is #4. The top-5 DVOA teams are currently the Eagles, Broncos, Seahawks, Cardinals and Vikings. The top-5 DSRS teams are the Seahawks, Broncos, Vikings, Patriots and Eagles.

 

[Jangles]

Stats are for losers

And this defense is poo

 

[luuked]

Aye, it's just an overview of what you can expect of a defense when they perform at their "usual" level. Also did it to see how the game situation affects the defensive play of different teams. Analysis that use a full game's data sometimes have a hard time distinguishing that.

In regard to DVOA, it's a well thought-out idea, but it's yardage-based. So I'm prefering to use it together with Defensive SRS, which is also designed to ensure comparability of different sets of data, which is based on points relative to average. So take the 2006 Patriots for example, #2 in total points allowed, #3 in points per drive. Their DSRS is #2 behind Baltimore. Meanwhile their DVOA is #7 because they finished #13 in yards per drive. They rank behind the 2006 Panthers for example who had a significantly worse points per drive figure (#11) but were #7 in yards per drive.

The Patriots DVOA is currently #19, dragged down by the horrific second half of week 2 against Miami (since week 3 the Patriots are #11 in DVOA) while their DSRS is #4. The top-5 DVOA teams are currently the Eagles, Broncos, Seahawks, Cardinals and Vikings. The top-5 DSRS teams are the Seahawks, Broncos, Vikings, Patriots and Eagles.

Thanks for the thoughtful response.

I definitely agree that DVOA being yardage based is one of th0se caveats. I am not that much into football statistics -- despite working as a data analyst -- mostly because team sports in general and american football in particular are almost impossible to model well.

However, your point was never to do any predicitive modelling or whatever so your tables work pretty well to give some pointers and facilitiate discussion.

 

[PF_SU]

mostly because team sports in general and american football in particular are almost impossible to model well.

Yes, everything that's "live" is hard to accurately and objectively model. The more fluid a game, the harder it is to model (I think that's one reason why there isn't much advanced analysis going on in soccer other than the expected goal and the goal impact metric). That's why I have as much of an issue with people who solely rely on statistics as I have with people who completely ignore statistics. Statistics provide context but don't replace the film. Hard data is never wrong though, what can be wrong is the interpretation of the data.

 

[patfanken]

Some comments on the comments:

1. Thanks for that awesome effort to put this all together, and more importantly putting some perspective on those numbers. These are the kind of threads that SHOULD be getting the most play from us.

2. What I find interesting (and please correct me if I'm reading them wrong) is that based on the how you parse the numbers you get wildly different evaluations of the Pats defense. In your second post you mention that the the Pats DVOA numbers put them below the midline at 19, however if you take their DSRS numbers they are 4th in the league. That is a pretty significant gap between the 2 evaluating systems.

3. All that leads me to question the entire statistical based evaluation technique. Doesn't it seem odd to you that you can use the raw numbers and come up with a system that puts the Pats as the #19th ranked defense, and a bad as people say, but when you use those same numbers using different priorities, the Pats are a top 5 defense. Proving the old adage that with a good accountant, you can make the books say pretty much anything you want them to say. The same with polls.

4. Now based on my viewing of the these numbers Pats is that they have a VERY strong back end (top 5). They are VERY good at stopping the run. (top 5), and they are not very good at rushing the passer. Whether that is by design or by the lack of personnel, has yet to be fully determined.

5. I say by design because historically BB has RARELY had individual pass rushers who got there by consistently beating players one on one. IIRC the BEST defenses of the early 2000's never had an individual pass rusher with over 10 sacks a season. It wasn't until Vrabel had 12 in 2007 did that happen. Even Willie McGinest never had 10 sacks in a season when BB was his HC. It makes me wonder if BB is even looking for a player with that kind of skill set, if there will be a negative at some other skills BB values as well.

Now that's not to say BB hasn't had defenses that ranked high in sack totals, they just seem to do it in a different way.

6. As far as the low number of sacks the Pats have had thus far. I wonder how far they'd rise in the rankings if they had just finished HALF the number of times they had the QB with both hands and couldn't bring them down. They must have had a dozen QB escapes allowed this season. I believe their sack number might even reach the "decent" level if they had managed to finish those rushes. In my opinion THAT's what they should be working on this off week, FINISHING their sack opportunities.

 

[deroc5050]

I didn't spend a lot of time but I didn't really understand what the numbers actually signify... but it looks like you put a lot of work into it. But with that said Belichick believes his defense is an issue and needs to get better. I would say they have an average D.. considering the offenses that they have played. I thought they would be a top 5 defense this year.. and that doesn't seem to be the case. But they usually get better as the season goes on.

 

[PF_SU]

2. What I find interesting (and please correct me if I'm reading them wrong) is that based on the how you parse the numbers you get wildly different evaluations of the Pats defense. In your second post you mention that the the Pats DVOA numbers put them below the midline at 19, however if you take their DSRS numbers they are 4th in the league. That is a pretty significant gap between the 2 evaluating systems.

The difference there is because DVOA is yardage-based, additionally factoring in down, distance, field position and score (and also the opponents' strength). Meanwhile DSRS is based on points allowed. Since Belichick defenses usually rank higher in points allowed than yards allowed they'll always rank higher in DSRS than in DVOA. The extreme difference of DVOA and DSRS right now is largely the result of the second halves of the Cardinals and Miami game. Especially Miami where they gave up tons of yardage and had practically no stops. If those two games are removed the Patriots' DVOA rises from #19 to #11.

4. Now based on my viewing of the these numbers Pats is that they have a VERY strong back end (top 5). They are VERY good at stopping the run. (top 5), and they are not very good at rushing the passer. Whether that is by design or by the lack of personnel, has yet to be fully determined.

In regard to the run, that's where DVOA actually aligns with the impression that the run defense is very good. The Pats' rush defense ranks #6 in DVOA behind Baltimore, Seattle, New York (Jets), Green Bay and Carolina. Considering that the Patriots in most games didn't key in on the run in the second half since they were ahead it's no stretch to say that they are top-5 material.

It makes me wonder if BB is even looking for a player with that kind of skill set, if there will be a negative at some other skills BB values as well.

I personally think it may be a cost/benefit thing. Belichick doesn't want to spend the huge amount of money that a pass rusher commands if he doesn't do everything else on a very high level as well. I'm sure Belichick would love to have Watt or Von Miller on the team who are great all-around players and a one-man wrecking crew, but he doesn't want to spend the money on a Mario Williams type of guy.

6. As far as the low number of sacks the Pats have had thus far. I wonder how far they'd rise in the rankings if they had just finished HALF the number of times they had the QB with both hands and couldn't bring them down. They must have had a dozen QB escapes allowed this season. I believe their sack number might even reach the "decent" level if they had managed to finish those rushes. In my opinion THAT's what they should be working on this off week, FINISHING their sack opportunities.

Yeah, I'd like to see a stat how many times they had the QB in their grasp only to let him go. Going by memory Taylor last week escaped some three times. Landry Jones also broke away. And in the week 4 matchup against Buffalo Sheard and Collins had both hands on Taylor on back-to-back plays but couldn't wrap him up turning a 2nd and 24 into a 1st down. They have to work on that and tackling in general. The amount of missed tackles this season is not something that Belichick wants to see.