Monday, January 3, 2011

Dubber: Effectively Measuring Offensive Efficiency

brilliant perspective and analysis courtesy of new author and face-melter, dubber

In "Developing an Offensive Gameplan", Brian Billick, using years of statistical data, identifies the four main areas critical to an offense's success (or attributable to its failure).

They are:
  1. Turnovers (pretty self-evident)
  2. Explosive Plays (defined as runs over 12 and passes over 15, given an even turnover margin, teams who garner at least 2 more explosive than their opponents will win about 80% of the time)
  3. Red Zone Efficiency
  4. Success on First Down (success defined as 4 or more yards). (The math behind this is a successful 1st down will populate a shorter 3rd down, which is easier to convert.)
I kept statistics concerning these four areas while watching the Seahawks and the Rams. This was an excellent game to apply this study to, because I generally feel the respective defenses of these two clubs are better than their offensive counterparts, and both punters, kickers, and coverage units had excellent games.

This puts the onus on the offensive gameplan and decision making.

As a side note, I have done this a couple of times now, and while the reports are getting better, I think keeping track of a couple other items (1st down run/pass ratio, and 3rd down conversions by distances) would be helpful. For the future.....

  1. Turnovers were even (one each)
  2. Explosive plays............few and far between, but they were correlated to scoring drives, notably the only touchdown in the game, which was set up courtesy of the game's biggest play (61 yard pass play)----------the Seahawks garnered this bomb, as well one more total explosive overall.
  3. 2 visits apiece, with the Seahawks coming away with 10 points to the Rams 6. The penalty in the red zone for the Rams was killer..........
It was probably these two categories that won the Seahawks the game.

I really didn't care who won this game, but of particular interest to me was the first down statistics, which bear out an important lesson from which coaches can learn.

4. First down

I took note of the yardage gained on every first down of the game. I computed an average. More importantly, I also took note of what happen from a conversion standpoint (IE, getting another first down) on first downs that garnered 4 or more yards, AND those that did not.

If I may put the cart before the horse, the Seahawks had 30 first down opportunities to the Rams 19.
  • The Seahawks averaged 5.5 yards on first down
  • The Rams averaged 4.2
While that tells the tale, it is also slightly misleading........obviously, whether the offense gains 13 or 50 yards on 1st down, they get another first down........and that 50 yarder can bolster my average enough to cover up my overall inefficiency on first down.

More important is to observe the RATIO in which I experience success on first down, not my average 1st down yardage.

  • Of the 30 times the Seahawks had a first down, they gained 4 or more a total of 17 times (57 %).
  • Of the 19 times the Rams had a first down, they gained to gain 4 or more a total of 9 times (53 %).
A big advantage for the Seahawks was their ability to overcome gaining less than 4 yards on first down.........they went ahead and converted 38% compared to the Rams 22%.

This was not due to a proficiency on 3rd down and long (in fact, both of these teams were generally sucky on 3rd and medium to long)..........a large part of this game was actually won and lost on SECOND DOWN!

The Rams (perhaps due to that first quarter bomb and the way the Seahawks were able to recognize and attack man coverage) played softer, zone coverage on 2nd down. Couple this with the Seahawks willingness to throw high percentage, risk averse passes on 1st and 2nd down (which is pretty ballsy when you are rolling with your backup quarterback), and you have a recipe for getting back into a manageable 3rd down situation.................
Here's the real kicker, and if you take nothing else from this, take this (remembering these teams are very poor offensively): Combined, the two team had 26 first down plays that gained 4 or more yards, and of those, only 6 failed to result in another first down.

That means that 77% of the time either of these teams gained 4 or more yards on first down, they ended up converting for another first down.

So, how is this helpful?

Go back to those couple of games where you felt like you didn't execute offensive, but should have. Not the game you blew a team out, and not the game where you were blown out........but that game that was close, or that you should have won, but your offense just struggled. Go back and look at just the yards gained on first down..................see what that tells you.

Charting first down is really the most time consuming part of this exercise. Turnovers are easy to count, as are red zone visits and explosive plays. Charting first down production means I must be completely focused on the television (something that rarely holds my full attention for hours on end).

I think 1st down run/pass ratio would be a huge addition, but I want it to be more structured than merely listing percentages........I would like to have some way to account for variance. For example, if I say a team threw 75% of the time on first down, that may be misleading if they throw a ton of screens. Also, there is a difference, in my mind, between taking a PA shot on first down, and taking a 3-step and throwing quick game. As a side note, while I believe the overarching theme of successful first downs should be unpredictability and balance, I would (personally) skew my first down gameplan more toward quick passing game than PA. I'd rather have the higher probability of 2nd and medium than risk an incompletion and leave 2nd and long.

Remember, an incompletion means you failed to gain 4 yards on first down, and are now off schedule...........and when the defense is better than you (a situation both offenses faced last night), you MUST stay on schedule.

Not that I wouldn't (and don't) take shots on first down, it just wouldn't figure predominantly into my general gameplanning practices.........

The type of PA also makes a difference........booting and throwing the comeback or the flat route is higher percentage than dropping straight back and throwing the NCAA route off run action.

And I understand, most PAP's have check downs, but I would like to have some way to delineate between taking a shot, and moving the pocket/still throwing high %.....which, given TV's horrible angles, would be hard.

At any rate, I would love anyone's thoughts on how to break this down.

Maybe the following 5 categories would work: quick/short passing, Verticals, PA's (maybe look at the difference between "going deep" and "high %"?), Runs, and Screens?
As I think about, perhaps screens could be sub-divided even further, as perimeter screens and slower developing slip or middle screens have entirely different functions for an offense.
Personally, a perimeter screen is like call sweep, and I consider it a safe way to get the ball in space (and get my 4 yards). Meanwhile, the latter mentioned screens are more like "home run" swings against a pressuring defense (lump them in with "PA shots").

3rd down

The really interesting thing Billick found about 3rd down is the conversion ratios for long, medium, and short were pretty standard. There wasn't a ton of deviation from the best to the worst offenses. For example, most teams convert about 80% of their 3rd and 1 situations, and convert a low percentage of 3rd and longs.
Doesn't matter if you are the Patriots or the Dolphins.

The difference?
The better offensive teams excelled at have MORE 3rd and short opportunities, while the bad offensive teams routinely faced drive killing (and turnover riddled) 3rd and long..........a direct result of good teams have first down success.

Still, it would be fun to chart that conversion ratio, maybe doing that a couple of times would reveal something about 3rd down philosophy.

As a final note, it was evident on 3rd down how much the Rams were still holding Bradford's hand...........their gameplan called for them to run only one formation on third and long (3x1 open with a compressed 2 and 3).........this kind of simplicity works just fine in high school, but in the NFL it's a different story.
I plan on taking some of the ideas in this second section and applying them when I do my next study (I'm thinking Colts/Jets), so if any of you have categorical or organizational suggestions, I'd love to hear them.


Dubber said...

I appreciate you posting this Brophy......

I hope the readers keep in mind this was cut and pasted directly from the forum. It comprises two seperate posts.

Brophy's only edit was the addition of cheerleaders, which is surely intended to distract the reader from the herky-jerky prose I used in the posts (if only the New York Times subscribed to the same school of journalism).

I will be doing this again (maybe the Colts-Jets game this Sunday), and will try to add some of the first down play selection analysis I discussed in the post.

I will either make that post "flow" more freely, or have Brophy add more pictures......


Anonymous said...

I don't understand how this is supposed to help. The point of Billick's guidelines is that, over time, those are relevant factors in deciding the outcome of football games, and further that they can be coached. Thus I could see someone studying the NFL season or the college season and showing how, in the aggregate, these factors did or did not explain who won and lost.

But picking one random game and chalking up how the teams did seems a little too narrowly minded. So what if the "wrong" team had won? Would Billick's factors be irrelevant? What would you do different as a coach? It's like flipping a coin twice and saying how it landed on heads....TWICE!

To do these kinds of studies you have to look at a larger data set. It's the sample size problem, as always.

brophy said...

I believe what Dubber is pointing out (and does a great job of it) is taking a game many watched against two relatively unspectacular teams (neutral matchup) and apply tenets of Billick's philosophy to it.

What Billick has done in his book was detail the contributing factors of successful / efficient offenses. He then goes on to break those offenses down through various relevant statistical categories. Using these clearly defined quantifiables, the game plan can be constructed for the staff.

How does it help? It narrows the focus for the coaching staff to define what EXACTLY should be run when - which plays and personnel groupings provide the highest rate of success to mitigate probable risk. Adding those up, series-by-series, the offense will find itself on the winning end of the margin.

Dubber said...


Certainly, a small sample size can lead to bad stats, however, I wasn't trying to take the stats from this one game and hold them up a "tried and trues" for all other games.

I compared this game to Billick's finding to see how his anaylsis holds up.

Billick's sample size was not small (encompassing years and 1,000 of games played).

It's like learning about Newton's Third law in physics class, and then going outside and throwing a rubber ball against a brick wall to observe it.

I realize "immutable physical law" is not the same as "statistical probability", but you get the idea.

It was an experiment.

Dubber said...

Brophy bestowed me with edit power, and I've attempted to tie the post together.

Should make a little more sense now.

Coach Hoover said...

A book that my football-nerdiness really enjoyed growing up was Football by the Numbers by Allen Barra. It showed the value of turnovers, giving -50 yds for an Int and -40 yds for a turnover. It then computed adjusted yards per play factoring in the turnovers. I used to do this for all the college teams to see who really had the best Offense and Defense, but 03 was the last yr.

What you can do quickly without crunching the #'s is look at the yards per play for two teams and then subtract a yard for each turover. When you do that, whichever team has the best yds/play wins about every time. As a matter of fact, this quick formula correctly picked the winner for every NFL game this weekend. Wow, I'm really procrastinating getting my grades done today, lol.

Sorry, not trying to get off topic, just trying to verify what dubber was reporting about turnovers and explosives (which contribute to yds/play) being so important.

Kevin said...

Is there a spreadsheet you can share that has some of this breakdown info on it?

As far as the first down classification of plays, why not simplify to run left right middle, pass short left right middle, medium...long...

I know this leaves out the difference between a 3 step slant and play action flat throw, but the targeted area to me would also be helpful when breaking down a game.

Dubber said...


I am looking for ways to make this better.....sorry, I do not have an excel sheet, but I will be breaking down another game over this weekend, and in that article I will include what I do use.....

I appreciate your suggestion, and this is my feel toward categorizing plays:

The area of the field being attacked is more of a situational thing (personnel matchups, front or coverage weakness, etc.)..........the FACT you called a run, pass, screen, etc. is what's of value, regardless of where you decided to attack.

At least from a whole gameplan standpoint.......

What you are talking about would be beneficial on a play-to-play basis........

Example: We are in I-Twins to the field and the defense is in an Over front with Cover 3 behind it.......we know we want to do X this time in situation Y, so where are the holes in this defense to accomplish that?

I'm more concern with the offense's decision to throw quick game on first down (general), rather than their decision to call slant/flat (specific).