FanPost

A pass-centric look at the coaching candidates

As the favorite topic these days is who would be a good coach, I’ll add my thoughts on the most frequently mentioned head coaching candidates.

As I’ve written before, I’m a believer in the theory that

you must pass well on offense and stop the pass well on defense or you’re going nowhere.

The importance of passing efficiency has been confirmed by multiple analysts. From Mathletics (Wayne Winston 2009, no link - book extact) …

Is it more important to have a good rushing attack or a good passing attack? Is rushing defense more important than passing defense? Is it true that turnovers kill you? During the early 1960s statistician Bud Goode studied what makes a team win. He found that passing yards per attempt (PY/A) on both offense and defense were the most important factors in predicting an NFL team’s success. This is intuitively satisfying because PY/A is more of a measure of efficiency than total yards passing. Since we divide by pass attempts, PY/A recognizes that passing plays use up a scarce resource (a down) … Using team statistics from the 2003–6 season we will run a regression to try to predict each team’s scoring margin (points for – points against) … The coefficients for offensive and defensive passing efficiency are almost triple the coefficients for offensive and defensive rushing efficiency. This is consistent with Goode’s finding that passing efficiency is the key driver of success in the NFL.

Brian Burke ran a regression and found the same results

A good passing game is far more important than a good running game in the NFL. It’s at least twice as important, and probably even more so. If we include interceptions as part of the passing game, passing efficiency and interception rates dwarf the importance of running efficiency by a factor of 4 to 1.

Looking at passing efficiency explains a large part of the success of an NFL football team. With the recent rules changes I think this is likely to become more pronounced. As such, I think that in any HC you would want to see a demonstrable ability to build teams that can pass the ball and stop the pass (or for OC an ability to pass and for a DC an ability to stop the pass). There’s obviously more to football, and it’s nice to have a good running game and a good run defense, but I would focus on the passing offense and pass defense first. This is an admittedly simple approach to looking at coaches but I’ll think you see that the results make sense. And if teams cannot perform well on this most fundamental level that’s a red flag.

Note on interceptions

The various regression models tend to include offensive interception rate and defensive interception rate. However, in the models interception rates are discounted because interception rates are inconsistent and hence limited in its usefulness in predicting future outcomes. Here's an explanation the importance of consistency.

How well a statistic predicts future outcomes is not just about the parlor game of picking winners. Stats that predict future outcomes measure the signal of how good a team really is, underneath all the noise of randomness … When trying to measure team strength or player ability, the focus should be on how well a team or player is likely to play in the future.

That's why stats that minimize noise, even while sacrificing some of the signal, can be more predictive and better measures of true ability than stats that include all the signal and all the noise.

A statistic that both correlates with winning and correlates with itself would be a reliable predictor of future wins. For example, team offensive net pass efficiency correlates with team win totals at 0.66, which is about as strong a correlation as you'll find. It is also fairly consistent throughout the season, correlating with itself at 0.55. The result of the two correlations is a 0.36 predictivity factor, which is the highest of the stats I measured.

The reason that I have not included an adjustment for interceptions is that different researchers keep finding the same result: a large portion of the variance in interceptions (up to 90%) can be explained by randomness. And for a statistic to a useful predictor we need to know that the results from this season can tell us what will happen next season. Interceptions fail on that measure. Brian Burke on interceptions.

One of the most random events in football is the interception. We can say that not just because of tipped balls or freak gusts of wind, but because we know that interception rates for teams and for individual quarterbacks varies widely from season to season. They also vary greatly from one half of the season to the next, which suggests that more is at play than player ability.

And here ...

What makes this model unique is that it distinguishes between factors that are explanatory (telling us why teams won past games) and factors that are truly predictive (telling us why teams will win future games). For example, turnovers and turnover differential account for a very large part of why teams won previous games. But turnovers are extremely random from week to week, and teams are not consistent from one part of the season to the next. They're obviously not entirely random. (Peyton Manning will never have a 4% interception rate--3% is league average), but a very large part of the variance in interception rate can be accounted for by randomness.

From Stumbling on Wins (Berri 2010) …

Why do teams track statistics in the first place? The primary purpose is to separate a player from his team. We know at the end of a contest who won. What we don’t know is which players were responsible for a team’s success (or failure) … What makes for a "useful" number? J. C. Bradbury2 argues that there are two criteria for evaluating a specific statistical measure. First, one must look at how the measure connects to current outcomes. Then, one must look at the consistency of the measure over time.

The first criterion can be illustrated by an examination of batting average and OPS (on-base percentage plus slugging percentage). Batting average is one of the oldest and most popular performance measures for a hitter in baseball. However, only 65% of the variation in a team’s runs scored can be explained by a team’s batting average. In contrast, OPS can explain 89% of the variation in runs scored. These results tell us that if you want to know who had the biggest impact on team success, a hitter’s OPS is a better indicator than batting average…

Connecting a statistic to outcomes certainly helps identify the "better" measures; but this is really only half the story. Statistics are used to make decisions about the future. Therefore, decision-makers need to know if what a player did last season says something about what he will do this season. In other words, one needs to consider the consistency of the measure. As Bradbury argued, a measure that’s consistent over time is probably measuring a skill. In contrast, inconsistent metrics are probably capturing luck or the impact of teammates. Consistency is measured by looking at how much of the variation in this year’s performance can be explained by what a player did last year….

For football, the problem is quite severe, especially with respect to turnovers. Interceptions and fumbles are often considered crucial to outcomes in football. Turnovers, though, are almost impossible to predict. For quarterbacks, less than 1% of the variation in a quarterback’s fumbles and interceptions can be explained by what the quarterback did the previous season.

Hence, I’m cautious about putting weight on a coach’s prior interception rate record.

What I have used is Pass YPA adjusted for sacks … which passes the smell test. If you look at different versions of Pass YPA Pass YPA adjusted for sacks has a nice relationship to Super Bowl winners (raw data here). That makes sense to me from a common sense point of view. While interceptions explain why teams won, a large part of the variance is randomness … ergo interceptions are less useful as predictors. And since these are the statistics for the regular season, in a sense they are predictors of the Super Bowl participants.

Bonus material on the randomness of interceptions if you’re not convinced ..

Pro Football Reference

It's pretty clear which performance category is least consistent from year to year and probably belongs to the individual quarterback the least. It's probably the one the general public uses to judge a quarterback the most--interceptions.

Wages of wins …

The problem facing decision-makers in the NFL is the numbers – which are often cited – don’t tell us very much about the future performance of a quarterback … when we turn to interceptions per attempt, explanatory power falls to less than 2% (these results come from an examination of 399 quarterbacks who played consecutive seasons from 1994 to 2007).

The Coaches

Wade Phillips

Out of curiosity, the first thing I did was look up the statistics for Wade Phillips.

One interesting note. Wade Phillips had the biggest difference between the unadjusted Pass YPA and the Pass YPA adjusted for sacks (i.e. Phillips defenses look much better with sacks are included). That isn’t true of all coaches … in some cases the rankings aren't much different … but it was noticeable for Phillips.

Phillips

Def. Rank Pass YPA

2002

Atl

14

2003

Atl

30

2004

SD

24

2005

SD

17

2006

SD

3

2007

Dallas

10

2008

Dallas

3

2009

Dallas

10

2010

Dallas

25

Bill Cowher

Cowher

Off. Rank Pass YPA

Def. Rank Pass YPA

2002

Pit

7

8

2003

Pit

19

18

2004

Pit

9

5

2005

Pit

3

3

2006

Pit

8

16

-The general consensus is that Cowher is the most attractive candidate. That is backed up by the numbers. Cowher’s teams had both strong passing offenses and strong passing defenses.

-One caveat: Pittsburg has continued to put up equivalent performance without Cowher which bring into question how much of the credit belongs to Cowher and how much credit belongs to the Pittsburg organization.

Jon Gruden

Gruden

Off. Rank Pass YPA

Def. Rank Pass YPA

2002

TB

22

1

2003

TB

7

5

2004

TB

17

2

2005

TB

25

12

2006

TB

29

28

2007

TB

14

3

2008

TB

14

11

-My nightmare pick. A supposed offensive genius who has produced an above average passing offense once during his 7 years in Tampa.

-Furthermore, we can see that Tampa’s success was driven by the defense, and the defensive success was due to Dungy and Monte Kiffin (we know that because Dungy has built successful pass defenses wherever he’s been).

-So you get offensive ineptitude without the defensive performance. I recommend passing on Gruden.

Jason Garrett

Garrett

Off. Rank Pass YPA

Off. Rank Pass YPA

2005

Miami

18

2006

Miami

18

2007

Dallas

2

2008

Dallas

11

2009

Dallas

6

2010

Dallas

4

-A mixed record. Some below average years in Miami and some very good years in Dallas

-Like Phillips, Garrett’s record is better when adjusted for sacks.

-I’d rather not see Garrett leave. His record is very Sean Payton-ey and we all know how that turned out.

John Fox

Fox

Off. Rank Pass YPA

Def. Rank Pass YPA

2002

Car

28

3

2003

Car

10

8

2004

Car

13

22

2005

Car

7

4

2006

Car

19

9

2007

Car

27

18

2008

Car

4

7

2009

Car

23

11

2010

Car

31

8

-I like this pick far more than I expected primarily because Fox has consistently produced above average passing defenses

-BTW, you can see the contrast in defensive personnel. Ever seen Carolina play? John Beason (their ILB) can get 40 yards downfield in pass coverage. I wouldn’t be surprised if Carolina can figure out how to use Jason Williams productively.

-If you remember, when Jerry hired Wade he said something to the effect of ‘I looked at the inputs the team needed and Dallas needed more defensive inputs’. I’m sure that Dallas needs ‘defensive inputs’ now.

-Fox is friends with Jerry so I could envision Fox being amenable to a scenario where Garrett stays as OC and Fox comes in as head coach. I think better results are likely with a Fox/Garrett pairing than Fox running the offense.

-Speaking of offense, that part of Fox’s resume terrifies me. He’s had some good years and some bad years passing the ball but he seems to be in love with the running game. Carolina has not one, but two all-pro caliber RBs with the result that they’re worse than Dallas.

1. The Carolina Panthers are the worst team in the league. The offensive line isn't anywhere close to where it has been in recent seasons. Even with a healthy DeAngelo Williams and/or Jonathan Stewart, the running game isn't good enough. The passing game flat out stinks.

In case you didn’t notice, GB had Brandon Jackson/John Kuhn at RB. You don’t need an elite RB to win. And you definitely don’t need to invest two first round picks on RBs to win. If everything else is in place, it’s a nice luxury, but don’t tie up too much cap space on RBs. Make sure you can pass and stop the pass first. On the flip side, if you build the run at the expense of the pass (e.g. spend a couple 1st round picks on RBs) you’ll probably be bad.

-If you don’t believe me, maybe you’ll find Football Outsiders more persuasive

This week, Arian Foster and Peyton Hillis finished atop our statistical rankings at running back. So the two best performances at their position came from an undrafted free agent and a seventh-round pick, respectively. Neither was expected to be his team's starting running back heading into the season, but it seems absurd now to consider the possibility that either might lose his job anytime soon.

Each year, the rushing leaderboards are littered with guys who didn't need a pedigree to get to the top of the charts. Among the top 20 rushers by DYAR this season are six players who were either drafted in the seventh round or went undrafted; there are only four such quarterbacks and three such receivers in the passing top 20.

That jibes with every bit of research we've ever done at Football Outsiders; whether it's rate of return on a draft pick, the effects of injury on an offense, or whether spending money produces a return, investing valuable resources into running backs isn't all that likely to return a feature back. And yet, despite the existence of players like Hillis and Foster -- guys available for peanuts during the offseason -- teams still pour millions into their backfield.

Perry Fewell

Fewell

Def. Rank Pass YPA

2006

Buf

6

2007

Buf

26

2008

Buf

19

2009

Buf

3

2010

NYG

1

-Fewell has produced some excellent passing defenses. And that’s not a fluke. Fewell is from the Lovie Smith (who himself is from the Dungy/Kiffin tree … notice a trend) coaching tree which has been very successful building passing defenses.

-Similar to a Fox/Garrett pairing, I think the combination of Garrett and Fewell could be successful.

Dom Capers

Capers

Off. Rank Pass YPA

Def. Rank Pass YPA

2002

Hou

32

12

2003

Hou

21

32

2004

Hou

18

26

2005

Hou

32

31

2006

Mia

n/a

7

2007

Mia

n/a

31

2008

NE

n/a

n/a

2009

GB

n/a

6

2010

GB

n/a

5

-This is a record that should give everyone pause. For a supposed ‘defensive’ coach, his record with Houston was downright ugly. Only the last 2 years with GB have been outstanding.

Leslie Frazier

Frazier

Def. Rank Pass YPA

2007

Min

17

2008

Min

8

2009

Min

12

2010

Min

22

-Frazier, also from the Dungy coaching tree, has managed to put together some defenses that are incredible against the run and average against the pass. You know what that produced? 8-8 seasons. I’ll pass.

Tony Dungy

Dungy

Off. Rank Pass YPA

Def. Rank Pass YPA

2002

Indy

4

11

2003

Indy

2

16

2004

Indy

1

21

2005

Indy

1

8

2006

Indy

1

14

2007

Indy

4

6

2008

Indy

9

9

-The best candidate in my eyes if Jerry could lure him out of retirement. Not only are the statistics as good as Cowher’s, you don’t have the uncertainty about whether it was the coach or the organization.

-Also has what the Cowboys need … pass defense.

Pasqualoni

Pasqualoni

Def

2008

Mia

16

2009

Mia

28

As an added bonus. Dallas’s new defensive coordinator performance in Miami. How do they pick these guys.

Another user-created commentary provided by a BTB reader.