K-Means Clustering of NBA Player Performance — 2000–2021

A data analysis project in R that explores NBA player performance through physical attribute trends, Oliver's Four Factors breakdown by position, and unsupervised K-Means clustering on the 2020/21 season.

---

🖼️ Visuals

K-Means Cluster Plot — NBA 2020/21

Three distinct player profiles emerge from the clustering of Oliver's Four Factors (eFG%, TOV%, ORB%, FTr%).

Elbow Method — Optimal Number of Clusters

---

📋 Overview

The project is structured in four stages:

1. Data cleaning — handling NA values, unit conversion (ft/inches → cm, lbs → kg), deduplication, and position standardisation.
2. Physical analysis — evolution of average height and weight across all NBA seasons (1950–2021), broken down by position.
3. Four Factors analysis — comparison of Dean Oliver's four offensive efficiency metrics (eFG%, TOV%, ORB%, FTr%) across the five positions.
4. K-Means clustering — unsupervised clustering of 2020/21 players based on the Four Factors, with the optimal number of clusters determined via the Elbow method.

---

📂 Repository Structure

ClusteringNBAperformance/
├── analysis.R              # Main analysis script
├── R/
│   └── helpers.R           # Utility functions
├── data/
│   ├── seasons_stats.csv   # Season-level player statistics (Kaggle)
│   ├── player_data.csv     # Player biographical data (Kaggle)
│   └── README.md           # Data description and provenance
├── output/
│   └── figures/
│       ├── kmeans_cluster.png
│       ├── elbow_plot.png
│       ├── four_factors.png
│       └── physique_scatter.png
└── docs/
    └── variable_glossary.md

---

📦 Dependencies

install.packages(c(
  "tidyverse", "ggplot2", "measurements", "factoextra", "GGally"
))

---

🚀 Reproducing the Analysis

1. Clone the repository:

   git clone https://github.com/marinoalfonso/ClusteringNBAperformance.git
   cd ClusteringNBAperformance

2. Install dependencies (see above).

3. Run the analysis:

   source("analysis.R")

The script will clean the data, produce all plots, and export figures to output/figures/.

---

📊 Data Source

Both datasets are sourced from the Kaggle dataset NBA Players Stats since 1950 by Gilermo:

File	Description	Rows
seasons_stats.csv	Season-level player statistics, 1950–2022	~24 000
player_data.csv	Player biographical data (height, weight, college, etc.)	~3 900

The datasets are included in this repository for reproducibility. Please refer to the Kaggle license for terms of use.

---

🔑 Key Findings

Stage	Main Result
Physical trends	Average NBA player height has remained stable (~200 cm) since the 1980s; weight has increased slightly, especially among Centres and Power Forwards.
Four Factors	Centres lead in ORB% and FTr%; Point Guards lead in eFG% and TOV%.
Clustering	3 clusters emerge: high-efficiency scorers (high eFG%, low TOV%), rebounders/interior players (high ORB%, high FTr%), and perimeter / transitional players.

---

👤 Author

Alfonso Marino
GitHub · Feel free to open an issue or submit a PR.

---

📄 License

This project is licensed under the MIT License.