The simplest way to be a great offensive basketball player is to take a lot of shots and make a lot of shots (see: James Harden). Of course, that is easier said than done. Getting your own shot is difficult and tiring, not to mention hunting exclusively for your own points can be selfish and not conducive to team basketball. And yet, we come back to what Harden is doing this season, more or less resetting the expectations for what is possible from a shots taken and made perspective. His absurd stat lines lead me to this article, where the goal is to examine the relationship between player usage and player efficiency, for both the NBA as a whole and the Sixers specifically.
Usage (USG): “[t]he percentage of team plays used by a player when he is on the floor”
True Shooting (TS): “[a] shooting percentage that factors in the value of 3-point field goals and free throws in addition to conventional two-point field goals”
We’ll use these two metrics (and some conventional box score stats) for the first section here looking at the NBA as a whole. All data is via NBA.com/stats, for the 2017-18 season and the 2018-19 season through 1/26.
Raw Usage and True Shooting
Figure 1 shows the usage rate and offensive efficiency metric for every player and every game with a minimum of 15 minutes. The highlighted 76ers are Joel Embiid, Ben Simmons, Jimmy Butler (Minnesota data as well), and JJ Redick.
Next, Figure 2 shows the mean of each highlighted player in comparison to All-Stars from the 2017-2018 season (except Andre Drummond and Draymond Green). The data for the players and All-Stars is from both the 2017 and 2018 seasons.
Moving on, we’re going to add a little complexity. Figure 3 takes the middle 90 percent of the true shooting values for each usage value and visually separates them from the more extreme 10 percent true shooting points.
What we see is that instead of some efficiency cliff after a certain usage, there is just a narrowing of the range. Part of this is obviously due to how missing one shot with a lower usage will impact true shooting more than missing one shot with a sky high usage. Still, there does not appear to be a high level trend that usage negatively impacts efficiency. In fact, there is technically an extremely slight positive trend. Figure 4 presents the same methodology, but for the pool of All-Stars and selected Sixers.
The shape of the frontier is more or less the same, with the obvious small usage games gone due to the high profile players in this dataset.
Usage Changes by Player
However, using raw usage and true shooting for those visualizations might not be the best choice. You can imagine that the distributions of each player with slightly different average usages would overlap, creating a uniform pattern - potentially hiding patterns. Figure 5 presents the scaled usage and true shooting for each player.
You’ll note that there are again some high and low true shooting artifacts when the usage figures trend too far below their mean. Figure 6 presents just the recent All-Stars and the four Sixers, in a hope to remove those low usage oddities.
The negative usage difference problems were removed, but the plot is mostly just a blob. Hard to tell anything at all. Figure 7 isolates just the four Sixers and their corresponding linear regressions and is rife for over-interpretation and hot-takery.
In order to isolate just the shooting part of usage, Figure 8 replaces usage rate with field goal attempts.
Let’s be crystal clear about what these regression lines mean. Across the board for the entire All-Star plus Ben and JJ group (for both scaled usage and field goal attempts), the maximum R-squared is 0.07, with the vast majority below 0.01. That means that on average, the x-variable (scaled usage or field goal attempts) explains less than one percent of the variability in the scaled true shooting. Therefore, these are pretty garbage for predicting true shooting based on either usage metric.
However, when we use field goal attempts as our usage metric, the models for most of our group of players are borderline significant. Consequently, the trend of the mean true shooting as a function of field goal attempts is potentially relevant. Of note, Ben Simmons, Victor Oladipo, and Kevin Durant have the most negative (significant) association between field goal attempts and true shooting. DeMar DeRozan, Draymond Green, and Kyle Lowry have the largest positive (significant) associations.
This was pretty interesting to me. I believed going in that there would be some usage and efficiency cliff. Anecdotally, there’s this belief that a certain type of player (non-shooter or limited athleticism) might have difficulty keeping efficiency high when usage spikes. The evidence presented here is far from a smoking gun, but does dissuade me slightly from that viewpoint. I only used essentially a year and a half of data because I mostly wanted to compare Ben to his contemporaries. Perhaps in the future, I can look at how the usage and efficiency relationship changes over the course of a player’s career. As always, if you think I missed something, let me know in the comments.