This article is part of our Numbers Game series.
Everyone knows basketball players play worse on the second night of a back-to-back set – the second of two games played on consecutive days.
Well, one of this column's two goals is to investigate whether commonplace assumptions are statistically justified, so this topic sounds like a perfect fit. Already this season we've determined that the significance of pace is typically overstated. This week, we'll look at back-to-backs and see if the actual production matches the narrative.
Setting Up – Making a fair test
We want an apples-to-apples comparison as much as possible. For example, early this season, John Wall sat out the second night of two back-to-back sets. Tomas Satoransky averaged 31.5 minutes in those two games -- he has only one game of more than 25 minutes the rest of the season. For Satoransky, the data would shows that he plays better during back-to-back sets, even though we know logically that this is misleading.
Here are some of the steps I took to minimize this kind of "noise" in my data. First, I used data from the entire 2014-15 season for this investigation (my 2015-16 database has been corrupted, and I did not have time to fix that before writing this). We want an entire season of data to ensure adequate sample size and diversity.
Second, I focused the investigation on players in the starting lineup on consecutive nights. Only players who started both games of the back-to-back set qualified. This doesn't eliminate all of the noise, but
Everyone knows basketball players play worse on the second night of a back-to-back set – the second of two games played on consecutive days.
Well, one of this column's two goals is to investigate whether commonplace assumptions are statistically justified, so this topic sounds like a perfect fit. Already this season we've determined that the significance of pace is typically overstated. This week, we'll look at back-to-backs and see if the actual production matches the narrative.
Setting Up – Making a fair test
We want an apples-to-apples comparison as much as possible. For example, early this season, John Wall sat out the second night of two back-to-back sets. Tomas Satoransky averaged 31.5 minutes in those two games -- he has only one game of more than 25 minutes the rest of the season. For Satoransky, the data would shows that he plays better during back-to-back sets, even though we know logically that this is misleading.
Here are some of the steps I took to minimize this kind of "noise" in my data. First, I used data from the entire 2014-15 season for this investigation (my 2015-16 database has been corrupted, and I did not have time to fix that before writing this). We want an entire season of data to ensure adequate sample size and diversity.
Second, I focused the investigation on players in the starting lineup on consecutive nights. Only players who started both games of the back-to-back set qualified. This doesn't eliminate all of the noise, but it does exclude Satoransky in the example above.
Third, I only compared players to their normal performance as a starter. In 2014-15, Aaron Brooks started 16 games, three of which were the second of a back-to-back set. Over the course of the season, Brooks also played less than 20 minutes in more than a third of his appearances. Most of those came during long stretches of low-minutes games, during which time he was off the fantasy radar. By focusing only on his performance in starts, we are limiting ourselves to the set of games most likely to impact fantasy teams. Additionally, Brooks' starts are more likely to have similar workloads than his season taken as a whole.
My final step is best explained through an example. Let's call any night that is not part of a back-to-back "A," the first night of a back-to-back "B," and the second night of a back-to-back "C." Brooks had 13 A or B starts and three C starts. If we compared his C starts to all 16 starts, then the three C games would be double counted. Therefore, I compared Brooks' – and every other player's -- performance in C starts to his performance in A and B starts, *not* to his average among all starts.
Setting Up – What are we looking for?
I compared players' minutes, points, rebounds, assists and field goal percentage. That, however, is a ton of data – way more than I could fit into one article. Therefore, as my baseline test of "are players actually worse on the second night of a back-to-back," I looked at total fantasy points. I used the DraftKings fantasy points algorithm, though any system should yield similar results -- FanDuel and Yahoo use similar algorithms.
I also paid close attention to minutes -- if players typically play fewer minutes on the second night of a back-to-back, then while that is an important finding, it will impact the reliability of the rest of our findings.
What did you find? - Minutes
First, I checked the minutes and the result was the most convenient result possible. On average, players played 0.2 minutes -- 12 seconds -- fewer on the second night of the back-to-back. For our purposes, that means we can assume players play the same amount time on either night of a back-to-back.
For the really high-minutes players, there is some nuance. The model showed that the really high usage players, those averaging more than 33 minutes per night, played slightly less on the second night. For players averaging 33 minutes, it was about a one minute difference; for players averaging 36 minutes, it was about a two minute difference. Still, this is a small enough discrepancy that we can comfortably ignore it.
No, really, what did you find? Like, about the point of this article?
Players play worse on the second night of a back-to-back. Wait! Don't go! Just because that is what we expected, there is still a lot of interesting stuff here! I promise!
Most important, it is better to know something than to assume it. Now we know, mathematically, that it is fair to expect players to play worse on the second night of a back-to-back. Well, at least to the math nerd, that's the most important thing. You all will probably care about the next stuff more, though.
Bigger role, bigger impact
The negative impact of a back-to-back is higher for high impact fantasy players. Again, this is not a shock, but it carries important fantasy value. There is little need to worry about the impact of a player on the second night of a back-to-back if the player averages only 20 minutes per game. While this is logical, it is easy to forget -- and it is something many analysts, myself included, can be prone to ignore.
Among the top 40 players (according to total fantasy points per game in non-second night of back-to-back situations), the decrease in production was about seven percent, roughly three fantasy points per game. In daily games, that can easily be the difference between winning and losing. In traditional fantasy, it is big enough to swing trade calculus.
As an example, ESPN's PR Averages ranks Kawhi Leonard as the ninth-best fantasy performer this season. A trade for No. 10 DeMarcus Cousins, therefore, seems basically fair. However, the Kings have 13 back-to-backs remaining, while the Spurs have only 10. Now, the impact of the back-to-backs is multiplied over several games. Perhaps that's not enough to kill the trade, but it does tip the scales in favor of the team getting Kawhi -- and plenty of fantasy managers have horror stories about losing leagues by fewer than 10 points.
Not all categories are impacted equally
This is the one area where my data was really surprising.
Top scorers see their points averages decrease by about five percent. Top passers see their assists averages decrease by about five percent.
Rebounds are often considered one of the "hustle stats," and so rebounds are definitely going to be the area most impacted by the second night of a back-to-back, right? Wrong.
Top rebounders see their rebounding averages decrease by barely two percent. Put differently, if someone averages 8.0 rebounds a game, over the course of seven normal games they would grab 56 rebounds. Over the course of seven second nights of back-to-backs, they would grab 55 rebounds.
Using the information
The better a fantasy player is, the more back-to-backs matter. Top 40 fantasy players were impacted almost twice as much as the league-wide average. As a general guideline, if a player averages less than 25 minutes, or if they are outside the fantasy top 80, the impact of a back-to-back is small enough that it can be mostly ignored. For those lesser fantasy players, over the long run, the second night of the back-to-backs will produce less -- but for any single night the impact may be imperceptible, and is likely to be outweighed by other factors.
For high-rebounding players, the second night of back-to-backs is less important. This is great news for all of the Enes Kanters of the world -- players whose value is dependent on only points and rebounds, with almost no expectation of value in any other category. Other examples players who play meaningful minutes, but due to their dependence on rebounds should be less impacted by back-to-backs include: DeAndre Jordan, Nikola Vucevic, Tyson Chandler, Tristan Thompson, Jonas Valanciunas, and Clint Capela.
In daily leagues, playing on the second night of a back-to-back is a legitimate knock against a player. Of course, all players should be considered in the context of their matchup and their recent performance. You should not be avoiding people only because of the back-to-back, but it is definitely a factor that is worthy of consideration.
In head-to-head leagues, back-to-backs are a valid consideration, especially late in the week in close matchups.
In roto leagues, the impacts of back-to-backs are best considered before draft day. For example, this season the Thunder had 13 back-to-backs, fewest in the league, while the Warriors have 17 and the Rockets have 16. Many managers had difficulty deciding who to pick first between Russell Westbrook, Steph Curry, Kevin Durant and James Harden. If a manager really thought they were effectively tied at the top, the significantly smaller number of back-to-backs would have been a worthwhile factor in Westbrook's favor.
