Does NFL Time of Possession Matter?
In the NFL, fans often focus of time of possession as an important indicator of how the game is going. The common consensus is that defenses tire more quickly than offenses. Therefore, the longer a defense is on the field, the more points they should give up.
At least, that’s the theory. But does the data support this time of possession theory? Are defenses that spend a long time on the field actually more likely to give up points? In this article we’re going to look at this question in a few different ways to put the theory to the test.
To receive email updates when new articles are posted, use the subscription form below!
Margin of Victory and Time of Possession
The first thing to check when verifying this claim is a simple plot of whether teams who have the ball longer tend to win more games. If time of possession matters, then winning the time of possession battle should help teams win more.
We plotted every game from the 2022 NFL season to see if this trend held up. On the x-axis is “home much longer the home team had the ball for than the away team”. Larger values mean the home team won the time of possession battle. On the y-axis is the home team’s margin of victory.
Immediately we see that this regression line has a noticeable positive slope. This means that teams who have the ball longer than their opponents tend to win by more. In fact, 100 extra seconds of time of possession equates to about a one point predicted margin of victory difference.
For example, if you had the ball 5 minutes longer than your opponent, then this would (all else held equal) translate to an expected margin of victory of about 3 points.
However, the next thing we notice is that there is still quite a bit of spread about the regression line. So, while time of possession does help us predict who will win, it is a fairly weak predictor.
More importantly, though, is that this does not really answer our question! This doesn’t tell us causation. All this tells us is that teams that have the ball longer win more often. Scoring drives are longer than non-scoring drives. Therefore, teams that score more points will have the ball for longer.
This tells us nothing about whether defenses tend to tire out during a game or not. It is still interesting that time of possession does correlate with winning. But it doesn’t let us know whether it is because defenses tire out. In the next section, we’ll look at how often offenses score as a function of how long they’ve had the ball for.
A Better Method
Instead of looking at margin of victory against who won the time of possession battle, we want to look at how scoring changes throughout the game. Remember that the claim is that defenses tire out faster than offenses. If this is true, then teams will tend to give up more points as the game goes on.
To test this, we looked at how scoring probabilities for drives change based on how long the opposing defense has been on the field for. The first plot below shows the probability of scoring a touchdown on a drive on the y-axis. The x-axis is how long the opposing defense has been on the field for.
This chart – while a bit noisy – shows that touchdown probability doesn’t really change much during the game. If defenses tiring was really something that mattered, we would expect to see a stronger trend, not the up and down pattern we see above.
There is one interesting feature in the above chart. UIp to about 1750 seconds (~30 mins of time on defense), the trend does begin to move upwards noticeably. In an average game, each team will have about 30 minutes of possession time. Perhaps this upwards trend indicates that teams start to perform worse once they are on the field more than 30 minutes which is more than they’re used to playing.
To say that another way, this data does seem to suggest that when teams play longer than average on defense, they start to allow more touchdowns.
Field Goal Probability Against Time of Possession
We want to repeat the above study by looking at field goal probability as time of possession increases. If defenses really do tire out as the game goes on, we would expect field goals to get more likely as the game goes on too. We repeated the same experiment as above and got the following data:
Aha! We see the same trend. Towards the end of the game, field goals do have an upward trend indicating, perhaps, that defenses tire.
But Wait….
Something fishy is going on. In both plots, we see an upward trend towards the end that could be misinterpreted as evidence that defenses tire as the game goes on. However, I claim that this effect is an artifact of the data and not evidence that defenses are tiring. Here is my evidence.
Point #1) The increases start at different times. In the TD Probability plot, the touchdowns start to increase at 1750s of possession. In the FG probability plot, the increase doesn’t happen until 2000 seconds of possession. If defenses really are tiring out, you would expect both these plots to show that it happens at the same time.
Point #2) If defenses really were getting tired, then the peak towards the end of the game should be higher than points in the middle of the game. The probability of touchdown should be largest when the defenses are the most tired. But, the peak at the end of the game isn’t higher than the rest of the data.
Point #3) The point at which the upward trend occurs depends on how we filter the data. In the plots above, we smoothed things out by averaging the data over a window of length 250. However, the plot below shows what happens when we average over a window of length only 10.
We still see the end-of-game peak here, but it doesn’t start until nearly 2300 seconds. This is inconsistent with what the data showed before! It was probably just a coincidence based on how we were filtering the data that caused this trend to start at 30 minutes of possession the first time.
The real reason that there is an upwards trend for touchdown probability at the end of games is because of the way overtime works. The really long games – practically the only ones that have one team with 2300+ seconds of possession by one team – almost always go to overtime. Scores do happen more often in overtime, likely because teams play more aggressively when they know they need to score a touchdown. Either way, this is the cause of the effect.
Intuition for Why Defenses Tire Faster
Now that we’ve shown that the data doesn’t support the claim that defenses tire out as games go on, we wanted to add a quick remark on why this would be the case.
At first glance, it might seem like teams should tire out at an equal rate. Every second the defense is on the field is a second that the offense is too. But in sports – like in war – asymmetric information can be used advantageously.
Every player on defense has to give 100% on every play because they don’t know what is going to happen. The offensive players can give less effort on certain plays especially when they know they are not going to be called upon. This means that defenses work harder on average when they are on the court. This is the intuition for why defenses should tire out faster.
Future Studies
Our study showed that there is no evidence behind the claim that defenses tend to tire out as the game goes on. While time of possession did correlate with winning, this was because scoring drives take more time than non scoring drives and scoring often correlates with winning.
In the future, we want to look at one more level behind this study. We used time of possession as our indicator of whether or not a defense should be tired. I think an overall better metric is to use real time instead of game-clock time.
For example, it is probably true that defenses will tire more quickly when they have to play lots of possessions in a row without much break. Without injuries or timeouts, the defenses will likely be less effective. I think it would be good to look at data correlating how fast the offense is playing with how many points the defense gives up. We could also look at expected points added to give the granularity necessary to conduct this study.
