[Paper Review] Offensive Lineup Analysis in Basketball with Clustering Players Based on Shooting Style and Offensive Role

Kazuhiro Yamada, Keisuke Fujii|arXiv (Cornell University)|Mar 4, 2024

Sports Performance and Training4 citations

TL;DR

This study introduces a novel framework for analyzing basketball offensive lineups by clustering players into distinct shooting styles using tracking data and offensive roles using annotated playtypes and advanced statistics. It applies Bayesian hierarchical models to quantify compatibility, revealing that combinations of All-rounder, Ball-handler, and 3-point Shooter players yield the highest scoring efficiency, while Bigs with mid-range or close-range shots show lower effectiveness.

ABSTRACT

In a basketball game, scoring efficiency holds significant importance due to the numerous offensive possessions per game. Enhancing scoring efficiency necessitates effective collaboration among players with diverse playing styles. In previous studies, basketball lineups have been analyzed, but their playing style compatibility has not been quantitatively examined. The purpose of this study is to analyze more specifically the impact of playing style compatibility on scoring efficiency, focusing only on offense. This study employs two methods to capture the playing styles of players on offense: shooting style clustering using tracking data, and offensive role clustering based on annotated playtypes and advanced statistics. For the former, interpretable hand-crafted shot features and Wasserstein distances between shooting style distributions were utilized. For the latter, soft clustering was applied to playtype data for the first time. Subsequently, based on the lineup information derived from these two clusterings, machine learning models Bayesian models that predict statistics representing scoring efficiency were trained and interpreted. These approaches provide insights into which combinations of five players tend to be effective and which combinations of two players tend to produce good effects.

Motivation & Objective

• To quantitatively assess the compatibility of player playing styles in basketball offenses.
• To improve interpretability by focusing exclusively on offensive features, avoiding confounding defensive attributes.
• To develop a dual clustering approach—shooting style from tracking data and offensive role from playtype statistics—for more precise lineup analysis.
• To use Bayesian modeling to estimate the impact of player pairings and lineup compositions on scoring efficiency.

Proposed method

• Shooting style clustering uses hand-crafted shot features (e.g., shot location, release time, dribble count) and Wasserstein distance to compare shot distribution similarity.
• Offensive role clustering applies soft clustering to playtype percentages and advanced statistics (e.g., true shooting, usage rate) to define composite offensive roles.
• Bayesian hierarchical models are trained to predict scoring efficiency metrics (e.g., points per 100 possessions) using cluster compositions as predictors.
• The model estimates both individual and pairwise effects of player clusters on offensive performance, enabling interpretation of synergy and compatibility.
• Interpretability is enhanced by using hand-crafted features and soft clustering, allowing direct mapping to intuitive player archetypes.
• The analysis is restricted to offensive data only, isolating the impact of offensive synergy from defensive or overall team effects.

Experimental results

Research questions

• RQ1 Which combinations of shooting styles lead to higher offensive efficiency in basketball lineups?
• RQ2 How does the compatibility of offensive roles—such as Isolation Attacker or Spot-up Shooter—affect team scoring output?
• RQ3 To what extent do synergistic pairings of players with complementary offensive roles enhance team performance?
• RQ4 Can soft clustering of playtype data effectively capture and quantify complex offensive roles beyond traditional positions?
• RQ5 How do the estimated effects of player cluster combinations compare to traditional positional clustering in predicting offensive success?

Key findings

• The combination of All-rounder, Ball-handler, and 3-point Shooter players produced the highest estimated effect on scoring efficiency, with a median effect of 0.3460 on points per 100 possessions.
• Lineups containing two strong scorers—specifically, Isolation Attacker paired with Ball-handler or 3-point Shooter—showed stronger positive effects than previously assumed, with a median effect of 0.2900.
• Combinations of Bigs who primarily shoot mid-range or close-range shots were consistently less effective, with the worst-performing pairings (e.g., Post-up Big and Stretch Big) showing negative effects, such as -0.4720 for PUB & WWH.
• The combination of Isolation Attacker with Wings, Ball-handlers, and Shooters was found to be highly effective, suggesting strong synergy between primary scorers and their supporting roles.
• Players with high shooting ability can partially compensate for less effective role combinations, as seen in SAS lineups where even suboptimal pairings did not drastically reduce scoring efficiency.
• The proposed clustering methods—especially the 10-cluster shooting style model—provided more interpretable and precise distinctions than prior work, such as Fan et al. (2023), due to the inclusion of dynamic tracking features.