While it's common knowledge (and a common complaint) that the NBA playoffs as a whole take a long time to complete, it's often the case that the games themselves take a long time as well. Part of this is due to overtimes being (slightly) more frequent in the playoffs than in the regular season—6.0 percent of playoff games go into overtime, compared to 5.6 percent of regular season games. Even if we focus on regulation length, though, playoff games, on average, last about 18 minutes longer than their regular season counterparts.
The graph below shows how this gap has varied over the years:
So what is driving these differences in game length? In a post on my site (subsequently shared on Deadspin), I examined how much "real time" elapsed for each minute of NBA game time. Not surprisingly, it's the last minute of regulation that takes the longest: 5.4 minutes on average for the 2013-14 NBA season.
My data source for that analysis was the backend files that feed the NBA.com play by play pages. If you know where to look, you can get a timestamp (down to the minute) for every play of every NBA game going back to the late nineties. This is a rich dataset which could be applied towards important topics like game flow and player rest, but let's go ahead and use it to settle some petty complaints regarding the length of playoff games.
What the hell is taking so long? As I mentioned above, every play has a timestamp attached to it. Using that data, I built a simple regression model which attempts to quantify how much real time each of the following game elements adds to the length of regulation:
- Timeouts (regular and 20-second)
- Free throws
- Non-shooting fouls
- Official reviews
I ran the regression separately for regular season games and playoff games. The results—how often each event occurred, how much time each instance contributed to the game length, and the total time contributed by the event (the first two figures, multiplied)—are summarized in the charts below:
Timeouts are the most obvious, and most maligned, contributor to increased game length. And while the number of timeouts only varies slightly between the playoffs and the regular season, the length clearly does. A regular season timeout adds about 2.1 minutes to the length of a game. A playoff timeout runs about 30 seconds longer than that. With 11 timeouts per game, that alone accounts for nearly 6 minutes of the 18 minute differential. And while I don't have any data to support this, it seems likely that the additional time is spent on longer commercial breaks.
To varying degrees, everything seems to take a little longer during the playoffs. Each free throw adds about 20 seconds to regular season games and 22 seconds in the playoffs. Substitutions add 9 seconds in the regular season versus 11 seconds in the playoffs. And not only do they take more time, playoff games feature more of these "dead time" plays: 2.6 more free throws on average and 1.7 more substitutions.
The chart below provides a (mostly) full accounting of that 18 minute time gap. As with timeouts, the breaks between quarters are longer during the playoffs, giving the networks up to 4.7 more minutes of ad time to sell (which translates to $4.7 million in ad revenue for each NBA Finals game, if the internet can be trusted).
At the bottom of the chart, you can see that there are three minutes of extra time that can't be explained by the regression model. I experimented with a few additional variables, but none were able to shrink that remaining gap. So, in the absence of statistical significance, I offer baseless speculation:
- Perhaps teams tend to contest in bounds plays more often during the playoffs, forcing the offense to use more of their allotted five seconds.
- Consistent with the slightly longer replay reviews, maybe officials are just generally more deliberative during the playoffs.
- Maybe the model needs a "Joey Crawford showboating" term.
I should also add that the inherent fuzziness of the data source may be contributing to the unexplained gap here. The timestamps only go down to the minute, so some additional massaging of the data was required, and that may be muddying the waters here. Regardless, I think we've probably applied more than enough analytical machinery to this particular topic. Feel free to add your own speculation about game length in the comments below.
Charts by Reuben Fischer-Baum