FanPost

Wed Morning QB: What I Liked, Didn’t Like, and a Leonard Davis size digression on the myth of run-pass balance.


What I liked

1. Zero sacks allowed: Dallas has now been sacked once in 3 games. I like to focus on what Dallas can control when evaluating performance. Sacks is one of those measures. If defenses can’t sack Romo that’s good for Dallas and real bad for opponents.

2. Sweet Play #1: On Dallas’s first series they ran a Romo bootleg with Bennett slipping into the flat. After going MIA last year, we saw this play run against the Skins with Austin & Witten rolling with Romo (Romo completed the pass to Austin). This week we saw it with Bennett. This used to be a staple of the offense and I’m glad to see its reappearance. Plus, Dallas used it to set up a play in the 2rd quarter when they ran a similar formation with Romo rolling to the right and throwing back across the field (which resulted in an incompletion to Bennett).

 

3. Special teams stopping the kick-off return to start the 2nd half inside of the 20.

4. Dez breaking tackles to get a first down. As posted in the comments on Monday.

Its unfair to Dez in some respects but its tough not to compare him to Michael Irvin. His possession ability and desire after the catch along with that #88 jersey are just too similar.

Creasy729

I agree with one addition. He’s more like Irvin 2.0. Dez has better speed and better ability to run after the catch. And I’m a huge huge Irvin fan. I wouldn’t say yet that he’s as good as Irvin at winning battles for the ball but he looks incredible so far.  

5. Sweet play #2: 5:36 3rd Quarter. Romo drops back, fakes the smoke screen to Dez, takes two more steps back and releases a perfect throw to Dez down the sideline for 30 yards. Here’s what Ioved about this play.

-It all happened in under 3 seconds. The play unfolds so quickly that Romo is able to execute the fake and release the ball before the defense had any chance to get pressure.

-I believe Garrett set this up in the Washington game.

As Jeremiah 24 commented after the Washington game

Just think if on one of those bubble screens Dez would’ve waited a split second (they were biting hard every time) for a pumpfake and then taken off it would’ve torched them. I guarantee you see it this season.

-This takes an incredible amount of skill to execute. I think there are maybe 5-6 teams in the league that could execute this play (Indy, NO, Atlanta, San Diego, New England, GB). This play is why it bothers me when people criticize Garrett. That type of skill is perfected through practice. Here's Garrett last week before the game.

We don’t use the word pressure a whole lot, we use the words prepare, approach

There’s no way a team executes like Dallas does without good practice habits. And make no mistake, Garrett has Dallas executing at a high level. I know it’s frustrating that Dallas can do the hard things and sometimes fails at the easy things. I think that’s better than doing the easy things but not being able to do the hard things.

Things I disliked:

1. The defense on Houston’s second series. The defense allowed 3rd down conversions on a 3rd and 7 and a 3rd & 19 (on a draw to Arian Foster … where’s the tackling Mike Jenkins!)

2. KO return unit. Another kick-off return that doesn’t make it to the 20 yard line. Dallas is currently 21st in kick-off returns.

3. Roy Williams reaching the ball out for extra yards when he already had a first down by 10 yards.

 4. 2nd defensive series of the 2nd half. Houston’s in a 2 & 17 and Mike Jenkins gets a penalty for illegal contact (and an automatic 1st down). Jenkins redeemed himself on the next play by getting an interception but it was still terrible to let Houston off the hook.

5. The Felix screen negated by Austin’s block in the back. Houston was blitzing and Garrett had the perfect play call. I’ll add here that I don’t think there’s really been a lot of difference in how Dallas has played over the first 3 weeks. Of all the major sports the outcomes in the NFL are subject to the most luck.

Exhibit 5 places sports along the continuum, using the skill/luck ratio based on the averages of the last five seasons for each sport. Appendix B shows the calculation for the NBA. Note that the length of the season is also an important factor: with only 16 games, the NFL has by far the fewest games and the NBA could have a much shorter season and still have a clear sense of which teams are best.

Skillversusluck_medium

via img38.imageshack.us


I know it’s boring to say ‘I don’t know. There’s a lot of luck in the NFL. This could just be randomness’ but that’s often true.

Here’s Football Outsifder’s real time commentary during the game.

Cowboys just killing themselves. The Texans blitz off both edges and Romo throws a perfect screen to Felix Jones that should have gone for a TD, but Miles Austin gets called for a block in the back.

6. On the first offensive series running Tashard Choice on 3rd & 2 (which he didn’t convert) and on the second series calling runs on 1st & 2nd down (which led to a 3 and out). I explain below why I didn’t like these calls.

APNY/A Update

For those of you following the ANPY/A statistics, Dallas won this game 10.8 to 3.4 (an outstanding performance for both the offense and the defense). For the season, the offensive ANPY/A is 7.2 (good for 7th overall) and the defensive ANPY/A allowed is 6.2 (24th overall).Statistics can be found here.

And now onto one Leonard Davis size digression.

Last week before the Houston game, in response to the concern about the offense which gained a lot of yards but hadn't scored that many points, I wrote "don't worry about the offense, they gain yards. Gain yards and the points will come". The only issue that is debated as much is "does Garrett call too many passes". Well, I'm here to tell you that the importance of run-pass balance is a probably a myth.

How many times have you heard "Dallas's offense is at its best when it's balanced", or "Dallas needs to run … between 2007-2009 Dallas is 1-6 when they run less than 20 times". What's been sorely lacking in this discussion is a review of the evidence. Is Dallas's offense really at its best when it balanced? What does the evidence actually say?

The evidence says that run-pass balance has very little to do with how good the offense is (as measured by offensive efficiency). I looked at Dallas's run-pass mix versus offensive efficiency (measured by rush YPA and pass YPA, I'll explain below why efficiency statistics are the correct measure to use) for every game from 2007-2009.

 

Below are charts of pass percentage (pass plays/total plays) versus rush YPA and pass percentage (pass plays/total plays) versus pass YPA. If there's a correlation we'll see a pattern in the data. If the correlation is positive (i.e. higher pass % = better efficiency) we’ll see the data moving from bottom left the top right. If the correlation is negative (i.e. higher pass % = worse efficiency) we’ll see the data moving from top left to bottom right.

53017126_medium

via img716.imageshack.us

33054366_medium

via img822.imageshack.us

As you can see the R squared is 0.06 and 0.16 for rush YPA and pass YPA respectively. (I didn’t include the chart but the correlation between pass % and interception rate is 0.0001 if you're curious). A couple observations.

-The correlations are very weak. A correlation of zero suggests there is no relationship. The correlation between pass percentage and rush YPA is effectively zero. To anyone who believes in the importance of run-pass mix this should give you pause. If the play mix is important how come there is no improvement in running efficiency as more passes are called? Wouldn't we see a little improvement in run efficiency because the defense anticipates pass? There’s just nothing there.

-It really doesn't matter whether you call 40% pass or 70% pass, there just isn't a large effect on offensive efficiency. This holds for both rushing efficiency and passing efficiency.

-While the correlation for rushing efficiency is effectively zero, there is a weak negative correlation for passing efficiency. That means that passing percentage and passing efficiency do move together. But we don't know which variable calls the tune and I believe this is where conventional analysis breaks down. Conventional analysis assumes that pass percentage calls the tune (i.e. that efficiency declines if you call too many passes). However, we know the key to winning (explained below) is passing efficiency. So here is an alternative explanation of causality. If a team is passing efficiently they're also probably winning -> leads to running the ball -> lower pass percentage. Or as Rabblerousr succinctly put it.

you must win to run the ball

Okay I said above that I’d explain why efficiency stats are the correct statistics to use. We’re at the point where you need to understand why so this all makes sense.

I've linked to this analysis from Advanded NFL Stats before but it's critical to understanding what's really going on. I strongly encourage everyone to read the whole article but here the key points.

-When you look at which statistics are correlated with winning passing efficiency is the most important

-passing efficiency allowed is the 2nd most important

-efficiency stats solve the problem of causality. No one thinks that a team winning made the team run or pass more efficiently (which patently doesn't make sense). Unlike rush attempts or pass attempts we can be confident that efficiency statistics are independent of the game situation (i.e. are not a result of who's winning). Having a 10 point lead doesn’t make a team pass for 10 yds per attempt but passing for 10 yds per attempt does make a team win.

Now I can calculate the correlation between pass efficiency and winning, and I will, but I don't think that's necessary to illustrate the supremacy of pass efficiency statisitics. Simpler is better and all you need to see is a table of the various statistics together and it's clear what the key is.

In table below I've taken scoring margin (positive scoring margins are wins, negative scoring margins are losses) and paired it with pass percentage, pass YPA, pass YPA allowed, and the difference between offensive pass YPA and pass YPA allowed. To make it easier to see the pattern I've sorted the data as well.

% Pass

Scoring Margin

 

Pass YPA

Scoring Margin

 

Pass YPA Allowed

Scoring Margin

 

YPA Difference

Scoring Margin

 38.7%

 9

 

 13.4

 10

 

 9.1

 (7)

 

 7.9

 21.0

 43.9%

 21

 

 13.0

 21

 

 8.7

 (2)

 

 6.4

 10.0

 45.5%

 10

 

 12.3

 13

 

 8.1

 (3)

 

 5.9

 28.0

 45.5%

 17

 

 10.4

 4

 

 7.6

 (21)

 

 5.7

 13.0

 45.9%

 1

 

 10.3

 10

 

 7.5

 13

 

 5.4

 18.0

 47.7%

 31

 

 10.0

 18

 

 7.5

 (38)

 

 5.1

 16.0

 49.1%

 4

 

 9.9

 6

 

 7.3

 (6)

 

 4.9

 10.0

 49.3%

 21

 

 9.7

 28

 

 7.0

 10

 

 4.8

 17.0

 50.0%

 (2)

 

 9.7

 25

 

 6.9

 (2)

 

 4.7

 6.0

 50.7%

 7

 

 9.6

 16

 

 6.7

 (2)

 

 4.3

 25.0

 50.7%

 28

 

 9.6

 17

 

 6.6

 (21)

 

 4.1

 4.0

 50.8%

 18

 

 8.7

 5

 

 6.6

 13

 

 3.7

 4.0

 51.7%

 10

 

 8.4

 11

 

 6.5

 17

 

 3.5

 17.0

 52.2%

 14

 

 8.4

 13

 

 6.5

 1

 

 3.4

 11.0

 52.5%

 16

 

 8.2

 24

 

 6.4

 5

 

 3.4

 24.0

 52.9%

 24

 

 7.9

 24

 

 6.4

 1

 

 3.3

 24.0

 53.2%

 7

 

 7.9

 7

 

 6.3

 4

 

 3.1

 12.0

 53.4%

 (3)

 

 7.7

 (3)

 

 6.3

 4

 

 3.0

 10.0

 53.7%

 11

 

 7.4

 4

 

 5.9

 (20)

 

 2.4

 7.0

 53.8%

 13

 

 7.3

 9

 

 5.7

 21

 

 2.3

 5.0

 55.4%

 17

 

 7.3

 (6)

 

 5.5

 14

 

 2.2

 9.0

 55.6%

 4

 

 7.3

 4

 

 5.5

 21

 

 1.8

 1.0

 56.4%

 (21)

 

 7.3

 17

 

 5.5

 7

 

 1.5

 31.0

 56.5%

 17

 

 6.9

 21

 

 5.4

 10

 

 1.3

 14.0

 56.7%

 21

 

 6.8

 14

 

 5.4

 25

 

 1.2

 21.0

 56.9%

 5

 

 6.6

 21

 

 5.2

 6

 

 1.1

 4.0

 57.4%

 6

 

 6.6

 (7)

 

 5.1

 21

 

 1.1

 21.0

 58.0%

 13

 

 6.5

 12

 

 5.1

 9

 

 0.9

 13.0

 58.1%

 4

 

 6.4

 (2)

 

 5.0

 11

 

 0.9

 (9.0)

 58.6%

 25

 

 6.2

 17

 

 4.8

 17

 

 0.7

 (4.0)

 59.2%

 10

 

 6.2

 1

 

 4.8

 24

 

 0.6

 (7.0)

 60.0%

 (7)

 

 6.0

 (21)

 

 4.7

 4

 

 0.4

 (10.0)

 60.3%

 24

 

 6.0

 10

 

 4.6

 24

 

    - 

 (6.0)

 61.8%

 12

 

 6.0

 1

 

 4.6

 18

 

    - 

 (20.0)

 62.3%

 4

 

 5.9

 (20)

 

 4.5

 7

 

 (0.3)

 17.0

 62.7%

 (20)

 

 5.5

 31

 

 4.5

 16

 

 (0.3)

 (2.0)

 64.4%

 (21)

 

 5.4

 1

 

 4.5

 (10)

 

 (0.4)

 (3.0)

 65.1%

 (6)

 

 5.1

 (2)

 

 4.4

 1

 

 (0.4)

 1.0

 65.3%

 (2)

 

 5.0

 (21)

 

 4.4

 (7)

 

 (0.5)

 (21.0)

 66.0%

 (21)

 

 5.0

 (7)

 

 4.1

 (9)

 

 (1.1)

 1.0

 69.1%

 (9)

 

 5.0

 (9)

 

 4.0

 31

 

 (1.4)

 7.0

 69.4%

 1

 

 4.9

 (10)

 

 4.0

 (4)

 

 (1.6)

 (21.0)

 70.3%

 (38)

 

 4.7

 (4)

 

 3.8

 28

 

 (1.6)

 (21.0)

 71.3%

 (7)

 

 4.7

 (38)

 

 3.8

 (21)

 

 (1.8)

 (2.0)

 72.7%

 (4)

 

 4.4

 (2)

 

 3.8

 17

 

 (1.9)

 4.0

 74.6%

 1

 

 3.3

 (21)

 

 3.6

 4

 

 (2.5)

 (7.0)

 75.9%

 (10)

 

 3.1

 7

 

 3.4

 12

 

 (2.8)

 (38.0)

 81.0%

 (2)

 

 2.8

 4

 

 3.0

 10

 

 (4.3)

 (2.0)

 

In the first two columns you can see pass % with scoring margin. A lot of the games with high pass % were indeed losses (remember that doesn’t show causality). However, there are losses sprinkled in the low pass % games and some wins sprinkled in the high pass % games. Next we have pass YPA with scoring margin. You can see that once pass YPA goes below 5 yards the outcomes are pretty bad (Dallas was 2-8 in those games). Next, is pass YPA allowed. At the high end the results are just as bad as the low end on the offensive side. Dallas is 2-6 in games where they allow a passing YPA of 7 yards or more. For both pass YPA and pass YPA allowed there’s a clear pattern but there are also results that go against the pattern (as you would expect as they only capture one half of the game, offense or defense). To capture both sides of the ball we combine offensive pass YPA and pass YPA allowed to get the pass YPA difference. Here’s the key observation. Dallas is 27-0 when the difference between pass YPA and pass YPA allowed is greater than 1 yard.

Here’s the same data presented with all 3 key metrics together so the pattern is easier to see.

Game

Pass %

YPA Difference

Scoring Margin

Game 4 - 2008

 81.0%

 (0.3)

 (2.0)

Game 9 - 2009

 75.9%

 0.4

 (10.0)

Game 13 - 2007

 74.6%

 (0.4)

 1.0

Game 14 - 2007

 72.7%

 0.7

 (4.0)

Game 12 - 2009

 71.3%

 (2.5)

 (7.0)

Game 16 - 2008

 70.3%

 (2.8)

 (38.0)

Game 5 - 2007

 69.4%

 1.8

 1.0

Game 15 - 2008

 69.1%

 0.9

 (9.0)

Game 6 - 2007

 66.0%

 (1.6)

 (21.0)

Game 4 - 2009

 65.3%

 (1.8)

 (2.0)

Game 6 - 2008

 65.1%

    - 

 (6.0)

Game 16 - 2007

 64.4%

 (1.6)

 (21.0)

Game 7 - 2008

 62.7%

    - 

 (20.0)

Game 8 - 2009

 62.3%

 1.1

 4.0

Game 14 - 2008

 61.8%

 3.1

 12.0

Game 3 - 2007

 60.3%

 3.3

 24.0

Game 13 - 2008

 60.0%

 0.6

 (7.0)

Game 7 - 2007

 59.2%

 3.0

 10.0

Game 12 - 2008

 58.6%

 4.3

 25.0

Game 8 - 2008

 58.1%

 (1.9)

 4.0

Game 11 - 2008

 58.0%

 0.9

 13.0

Game 5 - 2009

 57.4%

 4.7

 6.0

Game 10 - 2007

 56.9%

 2.3

 5.0

Game 7 - 2009

 56.7%

 1.1

 21.0

Game 15 - 2009

 56.5%

 3.5

 17.0

Game 9 - 2008

 56.4%

 (0.5)

 (21.0)

Game 2 - 2008

 55.6%

 4.1

 4.0

Game 11 - 2009

 55.4%

 4.8

 17.0

Game 1 - 2009

 53.8%

 5.7

 13.0

Game 9 - 2007

 53.7%

 3.4

 11.0

Game 13 - 2009

 53.4%

 (0.4)

 (3.0)

Game 15 - 2007

 53.2%

 (1.4)

 7.0

Game 16 - 2009

 52.9%

 3.4

 24.0

Game 6 - 2009

 52.5%

 5.1

 16.0

Game 3 - 2009

 52.2%

 1.3

 14.0

Game 12 - 2007

 51.7%

 4.9

 10.0

Game 1 - 2008

 50.8%

 5.4

 18.0

Game 4 - 2007

 50.7%

 5.9

 28.0

Game 14 - 2009

 50.7%

 2.4

 7.0

Game 2 - 2009

 50.0%

 (4.3)

 (2.0)

Game 3 - 2008

 49.3%

 1.2

 21.0

Game 10 - 2008

 49.1%

 3.7

 4.0

Game 11 - 2007

 47.7%

 1.5

 31.0

Game 10 - 2009

 45.9%

 (1.1)

 1.0

Game 1 - 2007

 45.5%

 6.4

 10.0

Game 2 - 2007

 45.5%

 (0.3)

 17.0

Game 8 - 2007

 43.9%

 7.9

 21.0

Game 5 - 2008

 38.7%

 2.2

 9.0

See it? That’s right, Dallas can pass the ball 70% of the time and win the game if they win the passing YPA battle. Dallas can also run the ball 50% of the time and lose the game if they lose the passing YPA battle. And all those games in the middle, where the play mix was 45-55% run/pass, Dallas can lose those too if they lose the passing YPA battle. What Dallas hasn’t done over the last 3 years is win the passing YPA battle by more than 1 yard and lose the game. Take a moment to let that sink in. Turn the ball over, get penalties, whatever, over the last 3 years as long as Dallas has won the pass YPA battle by more than 1 yard they’ve overcome everything else and won.  

I think the table is convincing. However, if you’re interested in the correlation here it is. The R squared between passing YPA difference and scoring margin is 0.41

17896640_medium

via img299.imageshack.us

Another comment that I see frequently is that Garrett is pass happy. Here’s the play mix data for the 2009 season for the entire league.  

Team

Pass Plays

Run Plays

Total Plays

% pass

ARI

621

360

981

 63.3%

SEA

651

386

1037

 62.8%

IND

615

366

981

 62.7%

CHI

598

371

969

 61.7%

PHI

592

371

963

 61.5%

DET

629

408

1037

 60.7%

SF

568

370

938

 60.6%

WAS

576

386

962

 59.9%

HOU

621

421

1042

 59.6%

STL

585

409

994

 58.9%

TB

560

396

956

 58.6%

GB

607

437

1044

 58.1%

PIT

586

427

1013

 57.8%

KC

583

427

1010

 57.7%

DEN

593

438

1031

 57.5%

DAL

584

435

1019

 57.3%

ATL

597

450

1047

 57.0%

NE

611

465

1076

 56.8%

NYG

574

440

1014

 56.6%

OAK

533

409

942

 56.6%

SD

544

427

971

 56.0%

JAC

563

445

1008

 55.9%

MIN

587

464

1051

 55.9%

NO

566

467

1033

 54.8%

BAL

547

462

1009

 54.2%

BUF

488

423

911

 53.6%

MIA

581

506

1087

 53.4%

CIN

546

523

1069

 51.1%

TEN

491

493

984

 49.9%

CAR

499

523

1022

 48.8%

CLE

471

495

966

 48.8%

NYJ

442

644

1086

 40.7%

Grand Total

18209

14044

32253

 56.5%

Dallas passed on 57.3% of its plays, 16th in the league. That’s right around the NFL average of 56.5%. The table below summarizes Dallas’s proportion of passes in Garrett’s 3 years as offensive coordinator.

 

% Run

% Pass

2007

 43.0%

 57.0%

2008

 41.0%

 59.0%

2009

 42.7%

 57.3%

I don’t see support for the hypothesis that Garrett is especially pass happy. Dallas is right around the middle of the NFL in terms of the proportion of their plays that are passes and that proportion has been relatively stable during Garrett’s tenure.

 

In fact, it’s possible that the offense would be more productive by calling more passes. Passing efficiency has improved since 2000 or so (average pass ANYP/A improved from around 5 ANPY/A to over 5.5 ANPY/A). In the past, when passing efficiency improves the run-pass balance shifts towards passing to take advantage of the improvement. However, the run-pass balance has yet to adjust to the increase in pass efficiency. That suggests that an offense could benefit by passing more and forcing the defense to adjust.I urge everyone to read the complete analysis from Advanced NFL Stats. 

Here's what I think is going on: Coaches took several years to fully take advantage of the increase in passing's effectiveness following 1978... The more recent and more subtle increase has yet to be realized. Passing, in most situations, has become more productive, but coaches haven't taken advantage of it yet.

Defenses appear to stubbornly focus on the run, making certain that they keep running efficiency under control. But this focus comes at the expense of passing efficiency…


Defenses may not have a choice. Perhaps once running efficiency gets much over 4 YPC, stopping an offense becomes extremely difficult. With 3 tries to get 10 yards, perhaps 4 YPC is a magic number that the basic rules of football dictate … defenses are unable to shift toward stopping passes more effectively…

 

If this is true, there is little doubt that offenses should take better advantage of the current imbalance by passing more often, specifically on first and second down outside the red zone. Offenses should force defenses to respect the pass more and more until they respond, thus allowing runs to become more productive.

I certainly believe this is true for Dallas. With Romo pulling the trigger and Austin, RW, and Dez receiving passing is a great strategy for Dallas even if it pushes the proportion of passes up to 60%, maybe even 65% (just a guess since Arizona threw 63% of the time). With the rules changes protecting QBs and WRs I believe there will be further increases in passing efficiency and we’ll see the proportion of passes rise. Coaches will realize that the best strategy is to call their best plays. For evidence of this, look no further than the Monday night game where both teams had less than 20 rushes and passed more than 75% of the time. I would add that it’s a bad idea to follow a sub-optimal strategy if your opponents are playing optimal strategy.

I know at least one person is going to say ‘sure, that’s all interesting but here’s where your analysis crumbles. If you pass on 1st down you end up in more 3rd and long situations. So while you might hit some big plays you won’t be able to sustain drives, and overall will be worse off.’ Except that’s not true. A while back I wanted to understand why Dallas had such a poor 3rd down conversion rate. It turns out that on average Dallas had the 31st longest 3rd downs in the league in 2009. I thought it might be play-calling, sacks, or penalties (it’s penalties BTW). As part of that research I looked at the proportion of passes on 1st down versus average 3rd down distance across the NFL(i.e. to see if passing on 1st down causes long 3rd downs). It turns out that teams that pass more of 1st down on average have shorter 3rd downs (the correlation is slightly negative and very weak). The result is what you would expect from the efficiency premium for passing (ANPY/A of 5.5 for passing versus YPA of 4.2 for running) and supports the hypothesis that teams would benefit from passing more. 

 

So the next time Tom Waddle says the key was that Dallas had a balanced attack, you'll know better.

As always, comments and criticism gratefully received.

 

Addendum:

 

If you’re interested in looking at the key metrics in spreadsheet format they can be found here.

 

 

Another user-created commentary provided by a BTB reader.

X
Log In Sign Up

forgot?
Log In Sign Up

Forgot password?

We'll email you a reset link.

If you signed up using a 3rd party account like Facebook or Twitter, please login with it instead.

Forgot password?

Try another email?

Almost done,

Join Blogging The Boys

You must be a member of Blogging The Boys to participate.

We have our own Community Guidelines at Blogging The Boys. You should read them.

Join Blogging The Boys

You must be a member of Blogging The Boys to participate.

We have our own Community Guidelines at Blogging The Boys. You should read them.

Spinner.vc97ec6e

Authenticating

Great!

Choose an available username to complete sign up.

In order to provide our users with a better overall experience, we ask for more information from Facebook when using it to login so that we can learn more about our audience and provide you with the best possible experience. We do not store specific user data and the sharing of it is not required to login with Facebook.

tracking_pixel_9341_tracker