≡ Menu

Attempting to measure fatigue in the NFL

Fatigue in the NFL is definitely real, and a team that’s tired is not a team that’s likely to excel. But I don’t know if it’s even possible to accurately measure the effect of fatigue in the NFL, and if it is, I certainly don’t know how to do it. Fatigue is a useful descriptive term but one hard to define. Is playing 3 games in 11 days likely to lead to a fatigued team? What about traveling west to east for a 1:00 game? How does that compare to being on the field for 10 minutes? And how does that compare to playing opposite a defense that’s gone 3 and out on three straight drives?

I don’t know. What I can do is look at the data we have from the last 12 years and see what general trends we can discern. So, are defenses worse off if they’ve been on the field for awhile?

There have been nearly 15,000 instances of teams having 1st and 10 near mid-field, defined as between the two 47 yard lines. On average, when teams gain possession in that area, they scored 2.2 points per drive. And, on average, those teams over the course of the season, averaged 1.75 points per drive over all drives.

So what happens if the “1st and 10 from the 47, 48, 49, 50, 49, 48, or 47” is the second play of the drive? Or the third? Or the 9th?

The 2.2 points per drive average when the situation occurs on the first play of the drive is the lowest in the group, although I don’t think that’s due to fatigue. Take a look:

Play #Pts/DrvAvg PPD

The middle column shows how many points, on average, teams scored in that situation, while the far right column shows the quality of the offenses in general (not that it really matters in this case). If fatigue had an impact in this situation, you would see the teams that start at their own 20, take 6 or 7 plays, and then have 1st and 10 at midfield be very successful. But that’s not the case.

It’s not surprising that the teams whose first play is the 1st and 10 at midfield fare the worst here, as those are the teams that haven’t proven anything; all other teams at least got one first down on the drive. The gambler’s fallacy would be to assume (correctly) that it’s really difficult to string together a long, 15-play drive, so teams on the 8th play on 1st and 10 at the 50 are unlikely to keep the success going. To be honest, I am at least a little surprised to see no discernible difference in outcomes. I also limited the data to just drives in the final 25 minutes of the game, but the results were largely the same.

I then looked at teams who started drives with 10-15 minutes left in the game, tied or trailing by no more than 8 points, and with possession between the 20 and 29-yard lines. I used four inputs — score margin, yards from the end zone, offensive team rating (on the season) and the number of additional plays run by the offense compared to the opposing offense.

As you might guess, only one of those variables was statistically significant: offensive team rating. The “yards from the end zone” was the second closest, and came close to mirroring the results we would expect. The “number of plays run” differential variable, however, had a p-value of 0.40, and was practically insignificant, anyway (a weight of -0.01). I tried using simply number of plays run by the offense instead of the play differential, but that was even less useful as a predictor.

I also looked at situations where a team had already run 30+ more plays than its opponent and then started a new drive. Excluding drives late in games (where a team might not be focused on scoring), on average, these teams started the drive with the ball at their 37 and scored 1.94 points. Additionally, on average, these teams were also leading 23-9.

What does that tell us? In general, you would project a team to score around 1.4 points with the ball on 1st and 10 at the 37, so that would be a sign that perhaps fatigue was playing a factor. On the other hand, these teams were clearly the better teams in these games (and these teams were generally better than average offenses), so that would cut against the idea of fatigue being the driving factor. The counterbalancing force would be that teams leading by 14 points aren’t as concerned with scoring, so that could artificially depress the results.

Measuring fatigue isn’t particularly easy. As Justice Stewart might say, you know it when you see it. In this post I looked at three different ways of measuring fatigue, but I’m sure you guys have more thoughts. I figured I’d throw this out to the crowd and see what creative ways you guys have for measuring the effect of fatigue.

  • Ben

    Dat GIF

    • Matt

      I love how the offensive lineman has NO idea what to do once he realizes Haynesworth ain’t getting back up

      • Independent George

        Man, that play never gets old.

  • Matt

    But not to take away from the overall quality of the post, great research. The only other factor I could see influencing fatigue and showing a result would be pace. If a team runs 30 more plays than a team it might be spread out over the whole game. But if a team has run 30 more plays in the same time of possession as the other team, perhaps then the defense would be tired and results might actually show?

  • Jim A

    Are you trying to measure offensive fatigue or defensive fatigue? Because the conventional wisdom is that defenses tire more quickly than offenses, but I’ve never seen any data supporting this. In fact, everything I’ve seen suggests neither side of the ball has any particular advantage resulting from fatigue-related factors.

    • Chase Stuart

      Trying to measure defensive fatigue, but yes, I agree it’s difficult to figure out how to get the data to measure this. Any thoughts?

      • Jim A

        I guess my point is that there’s no reason to think that offensive players don’t get fatigued also. The data shows that offenses don’t score more points as the game goes on, nor do they have any advantage at the end of a long drive. And being on the field a lot in a particular game doesn’t seem to have any predictive power other than the implication that the defense probably isn’t very good. So I would guess that any effects of defensive fatigue are probably negated by the effects of offensive fatigue.

        One idea might be to compare first quarter stats vs. fourth quarter (when the game is tied) or overtime stats. But even that can be tricky because late-game strategies are often different than early-game strategies. Matt’s idea to look at pace is an interesting idea, though.

  • George

    Fatigue is definitely a valid issue – e.g. the point from Winston about teams coming off a bye week typically playing about 2.something points better per game than expected. There was also a paper I’ve got that I will dig out, where a team from Swinburne University in Australia derived weights essentially for various elements in Australian Rules Football (AFL). They found the 2 hour mark on a flight was typically a threshold that made a difference of about 4 points from memory (e.g if an away team was flying more than 2 hours they were 4 points worse off – they did regression over about 20 years of data). The AFL is a lot higher scoring than the NFL but conservatively assuming an equal number of days off (which would be needed to rule out the short week issue), I think you could probably put a 3-5 hour flight being a 1 disadvantage to the team in question.