Accès libre

Two clusterings to capture basketball players’ shooting tendencies using tracking data: clustering of shooting styles and the shots themselves

 et   
02 mars 2025
À propos de cet article

Citez
Télécharger la couverture

Figure 1:

An overview of the proposed method
An overview of the proposed method

Figure 2:

Scatterplots on the feature space of shots whose features are reduced to two dimensions by UMAP; left: all shots, right: 3-pointers.
Scatterplots on the feature space of shots whose features are reduced to two dimensions by UMAP; left: all shots, right: 3-pointers.

Figure 3:

Mean Silhouette Coefficients in Clustering of Shooting Styles
Mean Silhouette Coefficients in Clustering of Shooting Styles

Figure 4:

Mean Silhouette Coefficients in Clustering of the Shots Themselves
Mean Silhouette Coefficients in Clustering of the Shots Themselves

Figure 5:

Distance between the shooter and the rim 3 seconds before shot in two clusters as examples. Normalized frequencies are displayed. On the left is Cluster 1, which tends to have the shortest distance from the rim 3 seconds before the shot. On the right is Cluster 2, which also tends to have a shorter distance from the rim 3 seconds before the shot.
Distance between the shooter and the rim 3 seconds before shot in two clusters as examples. Normalized frequencies are displayed. On the left is Cluster 1, which tends to have the shortest distance from the rim 3 seconds before the shot. On the right is Cluster 2, which also tends to have a shorter distance from the rim 3 seconds before the shot.

Figure 6:

2D histogram of shot location, in two clusters as examples. Normalized frequencies are displayed. On the left is Cluster 1, with shot locations concentrated near the rim. On the right is Cluster 2, which exhibits shots well distributed from the rim to mid-range.
2D histogram of shot location, in two clusters as examples. Normalized frequencies are displayed. On the left is Cluster 1, with shot locations concentrated near the rim. On the right is Cluster 2, which exhibits shots well distributed from the rim to mid-range.

Figure 7:

Visualization of shooting style clusters by t-SNE
Visualization of shooting style clusters by t-SNE

Figure 8:

Percentage of each shot cluster for Stephen Curry, Kevin Durant, Anthony Davis, and Dirk Nowitzki. Note that the total number of shot data is 216, 248, 280, and 262, respectively.
Percentage of each shot cluster for Stephen Curry, Kevin Durant, Anthony Davis, and Dirk Nowitzki. Note that the total number of shot data is 216, 248, 280, and 262, respectively.

Shooting Style Cluster Description

Cluster Name Description Mean Height [cm] Example Players
Close-Range Big Big Men who attempt most shots from close-range. 210.6 Andre Drummond
Dwight Howard
Mid-Range Big Big Men who can shoot from close-range as well as mid-range. 209.9 Pau Gasol
LaMarcus Aldridge
Mid-Range All-Rounder Players who play offense from mid-range and often shoot from near the high post. 208.9 Kevin Garnet
Dirk Nowitzki
Mid-Range Slasher Players who attack the rim from mid-range through post-play or drive. 206.5 Shaun Livingston
DeMarcus Cousins
Driving All-Rounder Players who begin offense from beyond the arc and aim to shoot from anywhere with their versatile skills. 198.6 Stephen Curry
Kevin Durant
Pull-up Ball-Handler Players who often drive and aim for many pull-up jumpers. 189.9 Kyle Lowry
Damian Lillard
Driving Ball-Handler Players who often drive from the perimeter, but also attempt threes moderately. 190.7 Russell Westbrook
James Harden
Stretch Four Big shooter who often attempts corner threes or threes from the top position. 205.2 Nikola Mirotic
Meyers Leonard
Corner Shooter Shooter attempting mainly three-pointers from the corner. 198.6 Patrick Beverley
Jason Terry
Pure Shooter Shooter who attempts 3-pointers from any location. 194.4 Eric Gordon
Kyle Korver
Slashing Finisher Players who do not shoot many threes and prefer to drive from beyond the arc. 199.9 DeMar DeRozan
Tony Parker
Driving Shooter Similar to Corner Shooter, but shooter with a slight preference for drive. 201.5 Klay Thompson
Vince Carter
Stretch All-Rounder Players who shoot from close-range to midrange, stretch and shoot 3-pointer as well. 205.7 Kristaps Porzingis
Kevin Love

Shot features and their units

Shot Feature Unit
x, y coordinates on the court of the shooter at the time of the shot meter
x, y coordinates on the court of the shooter 1 second before the shot meter
x, y coordinates on the court of the shooter at the time of receiving the ball meter
distance to the rim at the time of the shot meter
distance to the rim 0.5 seconds, 1 second, 1.5 seconds, 2 seconds, 2.5 seconds, and 3 seconds before the shot meter
distance to the rim when the ball was received meter
distance traveled while holding the ball meter
speed at the time of the shot meter / second
time of holding the ball second

Shot cluster names and examples of shots in each cluster_ The red and blue lines in the image represent the shooter's trajectory from 3 seconds before the shot to the time of the shot when holding the ball and when not holding the ball, respectively; the lighter the color, the later the time series_

Cluster Name Example Images
3-pointer from the right corner or right wing
3-pointer from the left corner
3-pointer from the top of the key or left wing
Mid-range shot near the high-post
Shot from a mid-post move
Close-range shot from a post move
Cutting layup/dunk
Mid-range shot from the right side
Mid-range shot from the left side
Driving layup/dunk

Top 5 players for Close-Range Big, Driving All-Rounder, Stretch All-Rounder, Slashing Finisher and Driving Shooter in TS%

Close-Range Big
Player Name (Team) FGA TS%
Hassan Whiteside (MIA) 682 62.9
DeAndre Jordan (LAC) 508 62.8
Cole Aldrich (LAC) 225 62.6
Andrew Bogut (GSW) 279 62.3
Steven Adams (OKC) 426 62.1