How Does Tempo Affect Offensive Production in the NFL?

There's an old phrase that people throw around to justify the purchases of oversized trucks, houses, TVs, and the like: "Bigger is always better." But do you prefer your Apple Watch that doubles as a smartphone, or would you like to go back to the cell phones of the 1990s that looked like a cinder block and weighed twice as much? Bigger is (not) always better.

What about the "new is always better" idea that a lot of people obsessed with progress say? Did you really prefer to play your Sega Dreamcast game console over your Super Nintendo? What would you say has more artistic merit: the classical composer Vivaldi's "The Four Seasons" suite, or Kesha's "Animal" album? Newer is (not) always better.

In the NFL today, also, there's a great perception that faster is always better for an offense. Philadelphia Eagles head coach Chip Kelly’s fetish with reaching 88 miles an hour on his offense has led people to believe that, if an offense is racing, they’re better. Is that true? Does up-tempo pace pump up the score, or will running an offense ragged run them into the ground?

Harder, Better, Faster, Stronger

To start this study, I compiled data from NFL offenses over the last five years (2010-2014) on how much value they produced on a season-by-season basis. To assess this, I used numberFire’s signature Net Expected Points (NEP) metric as the foundation of this piece. NEP helps us take the numbers we get from the box score and assign them contextual value so they relate even closer to the game on the field. By adding down-and-distance value, we can see just how much each play and each team as a whole influence the outcome of games. For more info on NEP, check out our glossary.

I paired this offensive value data with pace-of-play data in terms of offensive seconds per play (Sec/Play) –- compiled by Football Outsiders –- and total plays run. By taking each team’s annual rank in tempo and comparing it to their annual rank in Adjusted NEP (adjusted for strength of opponent) and Adjusted NEP on a per-play basis, I was able to see if there truly was a relationship between speed of offense and offensive value.

The table below shows the Pearson’s r correlation between each category of offensive pace and value. This is essentially a guideline of whether or not there could be a relationship, rather than an indication of cause-and-effect. Remember, the closer a correlation is to 1, the more likely a relationship is present; the closer to zero, the more unrelated the variables are.

Our guideline for understanding r correlation is that anything from 0 to 0.19 correlation –- positive or negative –- is a negligible relationship, .20 to .29 is a weak relationship, 0.30 to 0.39 is moderate, and 0.40 to 0.69 indicates a high relationship between variables.

Here we see that there is really no correlation between seconds per play and any sort of value produced. Both compared against Adjusted NEP and on a per-play basis, this falls into the “negligible” category. When it comes to total plays, however, there is an interesting relationship. Both in total Adjusted NEP and on a per-play basis, total plays reflects a positive correlation to value produced. This indicates some value in just being able to run more plays than your opponent, in the obvious accruing of more production. However, the positive correlation to per-play Adjusted NEP also indicates that there might be positive gain by wearing a defense down and gaining more efficiency on each play against them.

Our Work Is Never Over

So, this seems to debunk the myth of Chip Kelly’s “Blur” approach. Just for simple proof, the table below shows the fastest five offenses last year and their rankings in Adjusted NEP and per-play Adjusted NEP.

Playing fast is no guarantee of offensive success, but –- that being said –- it's interesting to note that last year was the only season in the last five where the fastest team in the league was not a top six offense in both total and per-play Adjusted NEP.

More Than Ever, Power After

All of this got me thinking about why per-play speed seems to have negligible correlation but the volume of plays run does. I also correlated average time of possession (T.O.P.) and Adjusted NEP to see if there was something to the old NFL adage that says “controlling the clock means you control the game”. The table below shows these results. Was T.O.P. a better indicator of offensive success?

This is the highest correlation between any sort of time measurement and offensive success yet. This, in fact, crosses into the threshold of “high correlation”, our highest bracket yet. Part of this correlation must be due to high-powered offenses getting ahead early on their opponents and then milking the clock to run out the game; remember, rushing attempts don’t stop the clock. Still, for this to be so consistent over a multiple year span of time and to match up with such a high correlation seems to indicate that there is more to it than clock killing.

So, it appears that faster is not always better, but (holding the ball) longer usually indicates a correlation to stronger.