Over the past decade there has been an explosion in data analytics in the NBA. On several occasions, data driven observers and athletes have butted heads over what is considered a “good” play and what is considered a “bad” play. One such point of contention has been the ever increasing number of 3-pointers attempted per game and the ever decreasing number of midrange 2-pointers attempted per game over the past several seasons.
A point made by the data analytics side is that the shooting trend is justified by the fact that 3-pointers generate far more points per shots (PPS) compared to midrange shots. For example, the average midrange shot is successful in 41.1% of attempts while the average 3-pointer is successful in 35.8% of attempts; 0.822 PPS for midrange shots and 1.074 PPS for 3-pointers.
As a counterpoint (as illustrated by the Twitter feud above), certain players suggest that open midrange shots are being passed up for contested 3-pointers. In this blog, we will investigate this suggestion and a more generalized idea: are there situations when “good” midrange shots are being passed up for “bad” 3-point shots?
TL;DR-KEY FINDINGS:
- Players are taking fewer midrange shots across the board regardless of “level of openness” (wide open, open, tightly guarded, very tightly guarded).
- Players are taking more 3-pointers while being tightly guarded, open, and wide open.
- Players are taking a similar amount of 3-pointers while being very tightly guarded.
- There is a subset of players (15% of active players, approx. 2 per team) who generate more PPS from midrange shots than 3-pointers. Of the shots these players take an increasing proportion are 3-point attempts while a decreasing proportion are midrange attempts.
- This subset of players are not becoming better 3-point shooters despite taking more 3-point shots and would generate more PPS if they were instructed to take a greater number of midrange shots.
Notes:
- Data was acquired through the NBA API (via the nba-api Python wrapper). The timeframe of this analysis begins with the 2013–14 season (the starting season when certain player positional data is available).
- The data is taken only from arenas with SportVU cameras (nearly every game played since the 2013–14 season).
- The 2019–20 NBA season was shortened due to Covid-19 so total attempts and makes by shot across the NBA will be affected.
- Midrange 2s are acquired from a dataset containing all shots at 10 or more feet from the basket.
- Closest defender distance during a shot has been binned into the following 4 categories by the NBA: very tightly — 0 to 2 feet, tightly — 2 to 4 feet, open — 4 to 6 feet, and wide open — 6+ feet.
- Ongoing project on GitHub, data analysis on Jupyter Notebook.
- Additional reading: David Foster Wallace essay — String Theory. Sometimes you have to give a little leeway to athletes while they are making a point.
Are there 3-pointers being taken when a midrange 2 would have been the better option?
We will first do some exploratory data analysis on general shot selection and closest defender distances.
As illustrated above, we can see that the number of midrange shots attempted has plummeted (nearly 50% fewer) while the number of 3-pointers attempted has skyrocketed (nearly 40% more) in recent seasons.
In order to investigate the idea that players may be taking fewer open midrange shots in order to take contested 3-pointers we must look at the amount of attempts made by a shooter’s “level of openness”.
We can see that midrange shot attempts have decreased across all levels of openness; meanwhile, the number of 3-pointers taken have generally increased and most conspicuously open to wide open 3-pointers. It appears that open midrange shots are possibly being passed up for open instead of contested 3-point shots.
So, we may need to dig a little deeper into the idea that there is an increase in “bad” 3-pointers taken at the expense of “good” midrange shots. For example, are there players who shouldn’t be taking as many 3-pointers?
Note: the number of wide open midrange shots will naturally be lower than the number of wide open 3-point shots since there are a greater number of defenders closer to the basket.
Who should be taking more midrange shots instead of 3-point shots and have those players been instructed to take the worse shot?
There are 494 NBA players who have taken both midrange shots and 3-pointers over the NBA seasons evaluated. Of these players, 73 (nearly 15% of players) generate more PPS from midrange shots than from 3-pointers.
As illustrated above, this subset of players have been increasingly taking a higher proportion of 3-pointers while a lower proportion of midrange shots. This lends credence to the idea that there are certain players who are being instructed to take an increasing number of “less than optimal” shots.
Are these players at least getting better at 3-pointers (with respect to PPS) as they take more 3-pointers?
We can see that there is a high level of variance with respect to 3-pointer PPS in this subset of players with much lower variance in their midrange PPS. Additionally, these players do not appear to be improving as 3-point shooters despite taking a higher proportion of 3-pointers. This further supports the argument that reversing the NBA’s general shooting trend for this subset of players would result in better shooting efficiency for a team.
Final Thoughts:
- On average a 3-point shot is a far better option than a midrange shot but certain players are more effective midrange shooters. By instructing these players to take more 3-pointers teams may be losing points.
- Further analysis needs to be performed on different subsets/clusters of players.
