Sharing is Caring?

An analytical look at the value of balanced offense distribution

Ball and One Sharing is Caring

When you think about a balanced offense, the first team that comes to mind is probably the San Antonio Spurs. Popovich’s beautiful passing offense truly feels like every player touches the rock and gets offensive opportunities. This well balanced sharing based offense has been not only well respected by many peer coaches, but has also produced one of the most consistently successful franchises in NBA history. With a resumé like that it should be easy to argue that the aim of every basketball team should be to have a balanced offense. But it’s also hard to deny the fact that having a superstar on the floor can also have some pretty profound effects on a team’s success.

Orgasmic ball movement
Orgasmic ball movement

Standard Deviation and Usage

Curiosity began getting the better of me, and I started thinking about how I could go about testing such a claim. One thought was to utilize the simple statistical measure of standard deviation. Standard deviation measures how much your data varies with respect to the mean. In principle, a balanced team like San Antonio should display data that has a lower standard deviation, whereas a superstar-centered team like Oklahoma City should display data that has a much higher standard deviation, where the data is more spread apart from the mean.

That's what we like to call a high standard deviation type of offense.
That’s what we like to call a high standard deviation type of offense.

But—as some of you keen readers might have wondered in this short paragraph—which data should I use?

Field Goal Attempts vs. Usage percentage

The next piece of the puzzle, was figuring out what stat I should use to illustrate offensive distribution and eventually apply a standard deviation analysis. The first stat that came to mind was Field Goal Attempts (FGA). Seeing how many players got a chance to shoot should give a sense of how the offense is being distributed. For example, if a team has 10 players with roughly the same number of FGA, you should have a lower standard deviation vs. a team that has a few players disproportionately shooting a ton. However, this isn’t exactly the best representation of offensive sharing.

Players can still be included in the offense even if they do not take too many shots, so that led me to using Usage percentage (USG) as my stat of choice. “USG is an estimate of the percentage of team plays used by a player while he was on the floor” and within its calculation it considers not only FGA, but free throw attempts, turnovers, and team minutes. The downside of this statistic is that if James Harden holds the ball for 22 seconds before passing to Trevor Ariza to shoot, it counts as though Ariza “used” the possession despite Ariza barely having the ball for a second. However, with a large enough data set, we should be able to ignore these one-off instances.

Gimme more shots bro
Gimme more shots bro

Data and Analysis

Good, now on to how we pull and prepare the data. With the good help of chief editor Jasper, we sourced our data from Basketball-Reference.com and decided to get all player box score game data for this past 2015-2016 season. *If you’re interested in obtaining a copy of this data, please email info@ballandone.com.* Next was cleaning the data for DNPs and aggregating the data. In order to aggregate, I grouped every single game by date and player team and took the standard deviations of USG for each game. I then averaged those standard deviations by team. Which led to the following results sorted by Average standard deviation of USG:

TeamAverage SD of USGSeason Rank
PHI7.02332530
BRK7.22541228
MEM7.68242915
DEN7.76639921
MIL7.89593922
CHI7.94845416
CHA7.9761439
DET8.06628512
DAL8.14528214
ORL8.19131620
NYK8.44766624
UTA8.45680119
BOS8.4975968
MIA8.53404410
PHO8.62822227
LAL8.6621629
SAC8.7367423
IND8.74388811
MIN8.81118626
SAS8.8340612
ATL8.8647677
TOR8.9985714
HOU9.20868717
WAS9.22242518
POR9.37011313
LAC9.4209166
NOP9.55250925
GSW10.025141
CLE10.1258713
OKC10.314815

Whoa…wait a second. The Sixers…Brooklyn have the lowest standard deviations? And jeez look at that glut on the higher end of the standard deviation scale, OKC, Cleveland, Golden State…

Hmm…So there definitely seems to be better teams the higher the standard deviations get. Let’s take a look using R 1.

sd-vs-season-ranking-chart

correlation1

Well that’s definitely something. So it looks like there is a decent negative correlation between season ranking and the average standard deviation of USG. In other words, the worse your team ranks, the more balanced your offensive distribution tends to be. Now of course this is only a correlation, and it doesn’t necessarily give us definitive proof of anything, but let’s think a little bit about the top five ranked teams displayed here.

TeamSeason RankUSG Rank
GSW128
SAS220
CLE329
TOR422
OKC530

Now the best of the best all rank pretty low in terms of USG distribution—the Spurs and Raptors showing a little more love than their top five counterparts—which means that they have a few players who dominate the ball more than others. This makes sense with the likes of Steph Curry, LeBron, Kyrie, KD, and Westbrook being the ballers they are, but I think the shocking thing about this is that it suggests AGAINST the idea that sharing the offensive load evenly is better than having a few players control the majority of the offense. Even San Antonio and Toronto don’t rank in the top ten for lowest standard deviations suggesting that maybe the optimal strategy is having a disproportionate amount of the offense controlled by a few of your players. To explore this further, let’s limit our sample to playoff caliber teams only and see what the correlation is between their ranking and their average standard deviations. 2

TeamUSG StdevSeason Rank
MEM7.68242915
CHI7.94845416
CHO7.9761439
DET8.06628512
DAL8.14528214
BOS8.4975968
MIA8.53404410
IND8.74388811
SAS8.8340612
ATL8.8647677
TOR8.9985714
HOU9.20868717
POR9.37011313
LAC9.4209166
GSW10.025141
CLE10.1258713
OKC10.314815

correlation2

Conclusion

Among the higher tier of the NBA the negative correlation between season rank and Average SD of USG is even greater. Now this shouldn’t have us jumping to any crazy conclusions like “damn we need to trade all of our players for a superstar ” 3, but I think it does highlight how important those stars are to teams with the right environment. For example, while Cleveland and Golden State illustrate a higher standard deviation, they are by no means stagnant offenses; however, they probably do benefit from concentrating their offense on their star pieces as opposed to letting everyone else get an equal share.

So at the end of the day, maybe teams do need a little Kobe in their lives.

Kobe knew balanced offenses were overrated all along.

footnotes

  1. Data handling and calculations done in R, representations in Excel
  2. CHI didn’t make the playoffs, but they did technically have a better record than Houston.
  3. hmm…screams Carmelo trade circa 2011
Ben Hwang

Author: Ben Hwang

Bball philosopher Ben Hwang is a NYC based consultant born and raised in NY who loves the Celtics...you may hate now.

Leave a Reply