NBA Players Need Their Rest

Jun 3, 2015; Oakland, CA, USA; Golden State Warriors center Andrew Bogut (12) sits on the bench during practice prior to the NBA Finals at Oracle Arena. Mandatory Credit: Kyle Terada-USA TODAY Sports
Jun 3, 2015; Oakland, CA, USA; Golden State Warriors center Andrew Bogut (12) sits on the bench during practice prior to the NBA Finals at Oracle Arena. Mandatory Credit: Kyle Terada-USA TODAY Sports /
facebooktwitterreddit
Jun 3, 2015; Oakland, CA, USA; Golden State Warriors center Andrew Bogut (12) sits on the bench during practice prior to the NBA Finals at Oracle Arena. Mandatory Credit: Kyle Terada-USA TODAY Sports
Jun 3, 2015; Oakland, CA, USA; Golden State Warriors center Andrew Bogut (12) sits on the bench during practice prior to the NBA Finals at Oracle Arena. Mandatory Credit: Kyle Terada-USA TODAY Sports /

In the past, I’ve looked at how Lebron James and Kevin Durant’s average speeds[1. As measured by the SportVU Player Tracking statistics] changed when playing on back-to-back days versus more days of rest. There was a clear difference and it turns out that it holds for the entire league:

days rest total large
days rest total large /

As we can see, there is a slight increase in average speed when players get more rest but the difference isn’t much (note the scale of the axis). We also see that at about 4+ days of rest, there’s a slight decline. But what’s interesting is that despite the small difference in average Speed between zero days of rest and one day of rest, the difference is statistically significant:

0 days vs. 1 day t-test
0 days vs. 1 day t-test /

However, as I’ve cautioned in the past, we have to be wary of assuming statistical significance when dealing with large samples. And of course, statistical significance is not the same as practical significance. So is this difference practically significant? On the one hand, the fact that players move slower on back-to-backs versus 1 day of rest makes complete sense but on the other hand, the average difference in speed is pretty small–one that would barely register as a blip if we were comparing two different players’ average speed[2. One possible way to correct for the sample size issue would be to sample 1000 observations from each group and see if the differences in means are statistically significant. We can do this exercise 1000 times and determine what percentage of t-tests showed as statistically significant. When running this little simulation, we find that in 14.8% of the t-tests showed as statistically significant at the 5% level while 24.2% showed as statistically significant at the 10% level. Both those percentages are fairly low and may indicate that in practical terms, the difference in means is negligible.].

However, another factor worth mentioning is that starters/high-minute players are the ones more likely to get tired on a back-to-backs versus a player who only plays 10 minutes a game. So rather than looking at all the players average speed, let’s look at only high minute players’ average speed by filtering out all the players who didn’t play at least 30 minutes in the current game and the previous game.

Days rest starters large
Days rest starters large /

The first thing you may notice is that the average speed is lower than in the previous graph (again, note the scale). This makes sense since players who are asked to play starter level minutes can’t maintain a fast average speed throughout the game or they will get too tired. Meanwhile, low minute players can go balls-to-the-wall for their limited minutes since they aren’t worried about maintaining their energy for a full game.

We also notice that the gap between zero days of rest and one day of rest has increased as we expected. At three days of rest, the average speed of a player seems to hit it’s height before going back down on 4+ days of rest (perhaps due to rust?).

So surely the difference between average speed on back-to-backs versus one day of rest must be statistically significant right?

0 days vs. 1 day t-test for starters
0 days vs. 1 day t-test for starters /

As we can see, it is a statistically significant difference. And unlike the previous t-test we ran, we don’t have nearly the sample size issues here. However, we can still perform the practical significance simulation test from earlier and we find that 66.3% of the t-tests showed as statistically significant at the 5% level while 83.4% of the t-tests showed as statistically significant at the 10% level. So what does this all mean?

The difference in means is still small and so it’s fair to still be skeptical that there is any difference. However, given what our intuition says, as well as past research, I would think the difference is significant. Players certainly aren’t a fan of the back-to-backs and there’s a reason for that: there’s less time for their body to recover and so it’s natural that they’re going to be playing and moving slower on the next night.

Of course, it’s also possible or in fact probable, that we aren’t capturing the true effects of playing on back-to-back nights. One other way we can look at this is to look at the decline in average speed from the previous game depending on the number of days of rest[1. Like earlier, filtered by 30+ minutes in the previous game and 30+ minutes in the current game.].

change in speed starters large
change in speed starters large /

We see that the decline in average speed on back-to-backs is even larger and that if players’ get at least one day of rest, their average speed improves in the following game. This trend continues until players get 4+ days of rest, where their average speed declines from the previous game, likely due to rust.

We can also look at the average distance traveled per minute based on the number of days of rest:

Average distance traveled large
Average distance traveled large /
Change in Distance large
Change in Distance large /

Like the previous graphs, we see the impact of the back-to-backs. Players are traveling more feet per minute on one or more days of rest until they reach 4+ days of rest, where we see the effects of rust potentially kick in.

Here’s a summary of the data in table form:

All Players

30+ minutes in current game and previous game

Of course, this is just the first step at evaluating how players perform following rest. In following posts, we’ll look at both team performance as well as looking at how metrics such as DRE, AWS and others change based on rest. Additionally, we’ll look at these metrics for older players (30+ years old) who presumably recover even slower and need that extra rest.

I’ll wrap the post up by looking at all the players and how they performed on back-to-backs versus more rest. While the data might not be presented in the best way, keep in mind you can search by Player.

All Players

All Players who played 30+ minutes in previous and current game