NBA Network Analysis
Project Overview
This project applies Network Science principles to sports analytics. By analyzing NBA data from 2020 to 2025, I modeled basketball teams as complex networks where:
- Nodes = Players.
- Edges = Assists (Weighted by frequency).
The goal was to go beyond simple statistics (points/rebounds) and understand the structural dynamics of a team: who is the real playmaker? How cohesive is the passing game?
Technical Workflow
The project is built in R and structured into reproducible notebooks (.Rmd):
1. Data Ingestion (API)
Instead of static CSVs, the project interacts dynamically with the BallDontLie API.
- Implemented secure API Key management via
config.R(git-ignored for security). - Used
httrandjsonliteto fetch and parse seasons, games, and play-by-play data.
2. Network Construction
This module transforms raw data into graph structures using igraph and tidygraph.
- Directed Graphs: Assists are directional (Player A -> Player B).
- Weighted Edges: The strength of the connection represents the volume of assists.
3. Centrality Analysis
This is the core scientific module (03_Centrality_Analysis.Rmd). I applied various centrality algorithms to determine player roles:
- Degree Centrality: To identify the main ball handlers.
- Betweenness Centrality: To find players who act as “bridges” between different sub-groups (e.g., passing from guards to centers).
- PageRank: To measure player influence within the passing flow.
Code Snippet (R)
Here is how the analysis pipeline is set up using the Tidyverse ecosystem:
# Setup dependencies and environment
source("setup.R")
# Load Network Analysis libraries
library(igraph)
library(tidygraph)
library(ggraph)
# Run the analysis notebook
rmarkdown::render("notebooks/NBA_Analysis.Rmd")
