This is the 3rd in a series of articles I am writing to analyze some common NFL statistics, focusing on how much value they have relative to team wins. I want to acknowledge the work of Brian Burke, Chase Stuart and even our own Matt Grecco, who inspired this analysis and whose methodologies I have leveraged, as well as Pro Football Reference, Armchair Analysis and NFL.com as the sources of my data.
There is beauty in the math of football. There are elegant formulas behind the physics of motion and force, and in the geometry of a football's trajectory. There are complex dynamical systems surrounding every fumble and probabilistic functions to be found in every match-up. All of these interactions create flowing, intricate patterns in measured data that we call statistics.
These statistics can be informative. They can crystalize the complexities of the game into meaningful narratives. They can deepen our understanding of a chaotic system. They can help us better appreciate the achievements of teams and players . . . or they can be Time of Possession.
Time of Possession (TOP) became an official stat in 1977 which means for 40 years, pundits have been jawing about controlling the clock and winning the time of possession battle.
Quite frankly, that kind of talk has always rubbed me the wrong way. The actual possession of the ball is irrelevant until you score points. Measuring how long you're not scoring points just seems un-American to me. It has the stink of a participation trophy.
Now don’t get me wrong. Being able to effectively slow the game down and possess the ball for long periods of time is extremely important. When you’re closing out a game or trying to keep a high-flying offense on the bench, “controlling the clock” increases your probability of victory.
No, what I’m talking about is the idea that old-school, smashmouth slow-paced offenses are more likely to win than teams that specialize in quick scores.
A while back, Chip Kelly tried to have the fastest offense in the NFL, and it failed miserably. Obviously, teams have to score touchdowns in order to consistently win, but it helps to wear teams down with long, punishing drives, and reap the rewards, particularly a rested defense.
As always, I will start with the airing of grievances . . . err, ranking of teams. Since Luck came to Indy, the Colts have had an average TOP of 30.7 minutes per game, which places them at a near-middle 13th place.
Since more TOP is supposed to be good then I guess that is not very encouraging. However, Lions and Texans and Chargers (oh my) round out the top 3 and I'm not feeling particularly envious of that group. So there's that.
Mapped out against win %, there is a definite signal that TOP and wins are linked. Below is the TOP correlation to win % and for comparison, I am also showing the win% correlations for total yards (a good metric) and passing yards (a bad metric).
Huh. TOP is actually more correlated to wins than total yards which is a pretty good metric.
If we go back to 2000 and look at how well teams played when they had 35-plus minutes TOP in non-OT regular season games, NFL teams compiled an impressive 83.7 winning percentage when they had the ball for that length of time.
Since 2000, there have been 86 games where one team had 40 or more minutes TOP in a non-OT game. Those teams were 80-6 (a .930 winning percentage).
So does TOP really matter? Should the Colts intentionally slow down their game to try and win TOP?
Well, only if TOP is actually causing those wins. For those who have read my previous posts, you know that I help determine that by measuring the predictive correlation(1).
(Darker shades = 16 game correlations, lighter tints = 8 game predictive correlations)
Aha! The predictive correlation drops below my completely arbitrary but nice and round threshold of 0.30 and lands at about the same level as the ever horrible passing yards, which means that while TOP is certainly tied to wins, it doesn't appear to cause those wins.
This issue is usually fleshed out when looking at the data by quarter and this is no exception.
I've seen this formula before: 3 quarters of “meh” + 1 quarter of “holy crap” = teams with a lead running out the clock. Of course TOP is tied to wins, the teams accumulating it have already won.
However, even if I am right (I am), the correlation in the first 3 quarters, while not great, is still respectable and so there is something linking early TOP to wins.
THE UNNECESSARY DETAIL
To get a better view, I split the data into winning and losing teams.
For the first 3 quarters, the eventual winner shows about a half a minute TOP edge on the loser. TOP can be deconstructed into two variables; the number of plays and the average time to run those plays.
If long, punishing drives win games and quick tempo offenses are miserable failures, then the winning teams should have longer average play duration than the losing teams.
Nope. There is no substantial difference in play length at all until the second half and specifically, the 4th quarter, which is likely due to situation (point differential).
Yep. Regardless of point differential, play length is pretty much the same until the 4th quarter at which point scoring doesn’t result from long play duration but rather long play duration results from not needing to score.
So, since TOP advantage isn't driven by play duration, it therefore must be driven by play quantity.
And that is what we see: winning teams have more plays which is driving more TOP.
This all is pretty straight forward. When teams play well, they get more yards per play, which extends drives more often and results in more points. That translates to leads which they can then protect while running out the clock. Don't believe me? Check out yards per play.
More yards, more plays, more points, grind clock.
Every aspect of that chain of events drives TOP without having anything to do with how slow or fast the team plays.
Winning TOP is a by-product of winning the game, not a cause of it. Chip Kelly didn't fail because his offense played fast: he failed because his offense played poorly.
This is usually where I impart my wisdom to you by giving you a better, related stat, but for this one I got nothin'.
Average play duration is 4th qtr situational as is yards per second. Points per second is good but it really is just a crappier version of points. Yards per second per play doesn't even make any sense but I tried it anyway and it's garbage.
I simply couldn't find, think of or invent any stat related to time of possession that was of any use at all. I feel bad . . . I feel like I've let you down . . . here's a picture of a kitten.
Measuring Time of Possession is worthless. Here endeth the summary.
1) Each team's 16 game season is randomly divided into two 8-game semi-seasons. The predictive correlation is the correlation between TOP from one semi-season to the win% from the other semi-season. The idea is that if TOP causes wins AND is repeatable then a team that is "good" at TOP in 8 games should replicate that and win in the other 8 games.