Shot Selection & Efficiency in the NBA Playoffs
Shot Selection & Efficiency in the NBA Playoffs
Strategic Shifts: The numbers behind playoff changes and their underlying causes.
Playoff basketball in the NBA is a different beast. Teams face off repeatedly, allowing for instrumental coaching adjustments. Only the league's best teams are playing, leading to more competitive games. There are countless avenues to explore comparing playoff to regular season basketball. In this analysis, I’ll focus on just one: shot selection.
Jump To:
Changes in Scoring due to Shot Location
The two ways to improve shooting efficiency are to improve shooting percentage or take higher-quality shots. Improvements in shooting percentage are relatively easy to quantify. Assessing the quality of a shot is a bit trickier1. In this analysis, I focus on the impact of shot selection2 on points per shot.
Consider three different points-per-shot metrics:3
Regular-Season (RS locations, RS Shooting %)
Playoffs (Playoff locations, Playoff Shooting %)
Expected Playoffs (Playoff locations, RS Shooting %)
Regular season and playoffs are true points-per-shot values. The expected playoffs metric shows what points per shot would be if players had maintained their regular season shooting percentages. By comparing the three metrics, you can calculate changes in points per shot due to shot location vs other factors (non-shot location).
Click this footnote for the underlying math4
When investigating the points per shot table, one thing stands out: change in scoring, due to shot location change, is very close to zero. At the median, different shot locations only account for 10% of the change in scoring. This supports the idea that most players are not taking a different mix of shots in the playoffs.
Key Examples
If shot location isn’t the driving factor for changes in playoff scoring, what is?
Alec Burks (DET & NYK): +0.473 points per shot
Alec Burks is a great example of both offensive scheme and volume. Last year, he played 43 games in Detroit, before being sent to the Knicks. In Detroit, Burks was a key contributor on offense5. In New York, he was more of a role player, with less scoring responsibility.
In the playoffs with the Knicks, Burks took a shot every 2.4 minutes. This compares to the regular season where he took a shot every 2.1 minutes6. Although this 20-second difference may not seem like a lot, Burks benefited in two major ways from playing with better teammates.
First, Burks could get more open looks on the Knicks due to his teammates garnering more defensive attention. Second, Burks’ role with the Knicks had less of a scoring focus. As a role player, he could focus on getting quality looks and let his talented teammates share the scoring burden. This might explain why he took fewer shots per minute.
Jalen Suggs (ORL): -0.180 points per shot
Suggs is a great example because although he shot from roughly the same locations, his points per shot decreased. This can be attributed to better defense or worse shooting percentage. I’m going to make the case that both are at play here.
Defense undoubtedly goes up a notch in the playoffs, with more talent and higher intensity. This might help explain Suggs’ dip in scoring efficiency.
The other prominent factor with Suggs is inexperience in the playoffs. 2024 was Suggs’ first-ever playoff series7, and playoff experience does matter8. Be it nerves, unfamiliar pressure, or anything else, this may contribute to the slightly worse scoring efficiency.
Changes in Shot Selection
Next, I am interested in changes in shot selection from the regular season to the playoffs. To do this, I cluster9 players based on shot location, irrespective of shot outcome. I cluster with three observations for each player:
Regular season 1 (all games)
Regular season 2 (against playoff teams)
Playoffs
I am interested in players who appear in the same cluster for Regular season 1 and Regular season 2, but in a different cluster for Playoffs. By including two regular season observations, I can control for defensive skill10.
Shot distributions by location for each cluster are as follows:
When investigating the above chart, some natural groupings appear. Think of these as areas of general areas of focus for a player, compared to their peers:
Restricted area (Cluster 0)
Above the break 3s (Cluster 1)
Corner 3s (Cluster 2)
Midrange & Paint (Cluster 3)
Out of the 86 players in the data, 18 (25%) were identified as having different shot locations in the playoffs11 :
Benefits of Changing Shot Selection
In game theory, a common occurrence is a Mixed-Strategy Nash Equilibrium (MSNE). MSNE is where players randomize their strategies, to maximize their payoff when playing against an opponent12. I believe that in basketball, shot selection is an MSNE. Players are incentivized to take shots from different locations on the court. This keeps the defense guessing and maximizes the shooter’s expected payoff13.
I also believe that the MSNE in the playoffs can change. To illustrate this, I’ll use the example of LeBron James. LeBron switched from a more perimeter-centric player to taking more shots in the restricted area in the playoffs. Why might LeBron drive to the basket more? Consider the benefits of each style of play.
Benefits of driving to the basket:
Less reliance on an open look
Higher chance of free throws awarded
Benefits of perimeter play:
Lower level of physical exertion
Lower injury risk
Even if the two approaches yielded the same points per shot, there are advantages to employing each.
In the playoffs, it might be harder to get an open three due to the improved defensive game plan and intensity14. Even with the same points per shot, this may decrease shots attempted, which in turn may decrease scoring. When driving to the basket, there’s also a higher likelihood of being awarded free throws.
In the regular season, teams don’t play multiple times in a row, and there is less emphasis on a specific game plan. This might let LeBron get more open looks15, increasing shooting volume (and overall points scored). Mitigating injury risk and physical exertion may also be of higher priority in the regular season. Appearing in more regular season games could outweigh the extra benefit of getting to the free-throw line.
Practical Applications & Next Steps
The biggest benefit of this approach is the framework for understanding where changes in shooting are coming from. Using this framework, with a few additions, one could build a great model for predicting changes in shooting efficiency. This would be a great addition to any team’s player evaluation process.
Incorporating player tracking data is a clear next step. Quantifying close-outs, contests, defensive rotations, and defensive intensity would be great additions. Eventually, instead of just two terms (location and non-location), a model could include 4 or 5 terms with each including a specific cause for change. There could also be an error term. You could assess the quality of this model by comparing this error term to the player’s volatility in shooting, and calibrate it accordingly.
Thanks for giving this article a read. For any methodology questions or to continue the conversation, reach out at vaughnhajra@gmail.com or on X/Twitter @vaughnhajra.
A true shot-quality metric would benefit from player tracking data, which is not publicly available.
Specifically shot locations and shooting percentages from those locations.
To calculate points per shot, multiply the shooting percentage of a location by the points value of the shot. These points per shot metrics exclude free throws, as the emphasis is on shot location and efficiency from those locations. I’m using 7 locations for these calculations:
Backcourt (3 pts)
Above the break 3 (3 pts)
Right corner 3 (3 pts)
Left Corner 3 (3 pts)
Mid-Range (2 pts)
Paint (non-RA) (2 pts)
Restricted Area (2 pts)
The difference between playoffs and regular season, is total change.
location and shot-values are consistent between playoffs and expected playoffs, so this difference is non-location change.
total change - non-location change = location change
In Detroit, Burks played 21.5 minutes per game. He was the only player to appear in 30 games and shoot above 40% from the three-point line.
In New York, Burks only played 13.5 minutes per game in the regular season.
Minutes per shot = 1 / shots per minute.
Burks took 8.3 shots per game in the playoffs and 8.6 shots per game in the regular season. He played 20.1 MPG in the playoffs and 18.4 MPG in the regular season.
This linked article may not be perfect (especially considering sample bias) but does a good job relating experience to playoff success.
Using K-Means clustering with 4 clusters. Data is standardized before clustering.
Note that defensive skill is distinct from defensive intensity. Increases in defensive intensity may lead to increased quality in the playoffs, which I discuss.
This doesn’t mean they abandoned their style of play. Rather they were likely somewhat similar to multiple clusters and shifted from one to another come the playoffs.
A simple MSNE example is a tennis player serving. If they only serve their opponent’s weaker side, their opponent will expect it and be better prepared for a return. If they serve to their opponent’s weakness most of the time, and randomly switch to serving to the strong side, their opponent will be left guessing, and at a relative disadvantage.
If this weren’t the case, players would only shoot from their highest percentage look. You could make the case that a select few players have a dominant strategy to only shoot from one spot (bigs in the post), but even these players will take one or two shots elsewhere.
Recall that this change is not due to defender’s skill, as I control for that using regular-season: against playoff teams. There may still be increases in intensity, though, which could contribute to this.
This is considering a case when a player will take a “good shot”, and not take a “bad shot”. What a player considers a “good shot” may slightly change in the playoffs, but likely not enough to change this idea.