Chip Kelly taught us that time of possession is not meaningless

If my web searching skills are any good, these were the first sentences written by Mark Saltveit for BGN three years ago:  "If you want to piss off Chip Kelly, it's easy. Just ask him if he's worried about his team's time of possession."   Naturally, Kelly was asked and he brusquely dismissed any notion that time of possession has even a small role in a game's outcome.  His response:

"Two years ago, a Thursday night game, we played UCLA. They ran 73 snaps, we ran 71 snaps. They had the ball for 40 minutes, we had the ball for 20 minutes. We won 60-13."

To Kelly, the media are nothing but the buzzing of flies, and time of possession is nothing but the byproduct of efficiency.  It's the third derivative of success, the scrapple of football metrics.  His teams had always possessed the football less than his opponents.  It didn't matter, though, since they usually won… in college.  However, "controlling the clock" is viewed in the NFL more commonly, sacredly even, as a sound strategy for success.  Now that Kelly is no longer head coach of the Philadelphia Eagles, is time of possession as meaningless a statistic as we were led to believe?

I initially thought so.  In response to Saltveit's article, I wrote here that time of possession was indeed meaningless, especially when one considers efficiency:

"Think of offensive and defensive efficiency this way… If both Eagles’ units, offense and defense, were abnormally efficient, if the offense scored on the first play of each possession and the defense secured a turnover on every single play, then time of possession and total plays for each team would be equal, yet the score would obviously be lopsided. This is why Chip Kelly dismisses time of possession and does not just preach total plays. He also preaches efficiency, which the media at-large seems to grossly minimize."

For the past several years, I've considered this as gospel: time of possession is meaningless!  Chip Kelly's philosophy merely reinforced my own conviction.  Admittedly, it's quite a romantic idea.  There is something innately attractive about doing what you consider is "right" while spitting in the face of convention.  Because in a vacuum this makes sense.  If an offense scores on every possession, who cares if it did so on a two minute drive or an eight minute drive?  If a defense consistently prevents scores, who cares if it's after a three minute drive or a six minute drive? Chip didn't, because for years he coached successfully in the NCAA vacuum.  But in the NFL, where parity is the rule rather than the exception, no such vacuum exists. (By the way, the UCLA team that Oregon so handily beat 60-13 finished the 2011 season with a 4-8 record.)

When looking at the 256 regular season games this past season, winning teams possessed the ball just 3.6 minutes longer on average than losing teams.  Translated in terms of win probability, every one minute  advantage over an opponent increased a team's chances of winning by 8.4% (click here for full logistic regression results).  Given all of the normal caveats about causation and correlation, let's not attribute this to strategy, protecting leads, running out the clock, etc.  Let's assume instead that 3.6 minutes and an 8.4% win probability serve as baselines.  These are typical byproducts of an NFL game.

The question though is ultimately about strategy.  For example, what if the Eagles were to play a team like the Carolina Panthers, whose offense is better than the Eagles' defense?  According to NFL convention, a sound strategy may involve the Eagles' offense slowing down the game in order to limit the amount of time Carolina possesses the ball.  This is the type of strategy Chip Kelly would never consider.  He'd more likely put the onus on the Eagles defense to limit Carolina's opportunities than have the offense do anything outside of his conceptual philosophy.  But, in this scenario, what does NFL convention look like?

During the 2015-16 season, there were 125 games in which a team's defense faced an opposing offense whose DVOA rank was at least 10 positions better.  If time of possession were meaningless, we might expect that the baselines above would pervade.  Instead, teams with weaker defenses that won possessed the ball 4.6 minutes longer on average than teams with weaker defenses that lost, and 6 minutes longer than the opponents with better offenses.  Every one minute advantage in time of possession translated to a 15.6% increase in win probability.  This is almost double the advantage of the "typical" NFL game.

Average NFL time of possession and related win probabilities for 2015-16 season pic.twitter.com/q0ketWV3vG

— Whoever THAT is (@JeromesFriend) February 27, 2016

What if the opposite were true?  Given what we know from above, we might expect that time of possession would matter less if a team's defense, in turn, is at least 10 DVOA rank positions better than the opponent's offense.  And that's what happens.  In this scenario (123 games), winning teams with stronger defenses possessed the ball 2.2 minutes longer on average than losing teams with stronger defenses, but this only translated to a 5.8% increase in win probability per one minute advantage.  Notably, this is below the established baseline.

This seems wonderfully intuitive.  Teams facing a stronger offensive opponent benefit from a ball control strategy more than teams facing a lesser offensive opponent.  Obviously this is a generalization, but it's a significant one.  There will always be games that demonstrate how time of possession may not matter.  For example, take Philadelphia vs. New England.  The Eagles only possessed the ball for 25:44 which, given the DVOA ranks of the units for each team, only afforded the Eagles an 11% chance of victory.  Yet, thanks to defensive and special teams touchdowns, the Eagles escaped with a win.

Hopefully, this time of possession/DVOA model can help explain these anomalies.  In week 9, for example, the Eagles defeated the Dallas Cowboys in overtime, but only possessed the ball for 25:42.  Because the Eagles' defense was 14 DVOA rank positions better than the Cowboy's offense, ball control did not matter as much.  It's under this type of condition that Chip Kelly is most comfortable coaching.  Since his defense was statistically better than his opponent's offense, he was freely able to implement his high tempo/less time philosophy and still have a 58% chance of victory.

What I once considered as gospel no longer seems to be true.  I will no longer say that time of possession is meaningless.  Instead, it's situationally meaningful.  There are certain conditions under which a ball control strategy proves to be advantageous.  As it turns out, duh.

So Mark… The next time the 49ers play Seattle or Arizona, ask Chip if he's worried about time of possession.  And tell him it matters.  I'll send scrapple.