peasant98 / NBA-Stats-Clustering

A repo with various tools and tricks for clustering of NBA Players.

NBA Stats Clustering

A repo with various tools and tricks for clustering of NBA Players.

Includes:

• Gaussian Mixture Models

• K-Means Clustering

• Hierarchical Agglomerative Clustering

• PCA

• Yearly Stats

Requires

pandas
numpy
nba-api
argparse
scipy
mpl_toolkits
matplotlib
sklearn

Also, needs Python3.6 or above due to f-strings.

Usage

git clone https://github.com/peasant98/NBA-Stats-Clustering/
cd NBA-Stats-Clustering/
python3.6 main.py

This will run the file and make a bunch of plots, which will be stored in the img/ folder.

• The above command will run clustering on points, rebounds, and assists for this season from k=3 to 9 on k-means (3 different initialization methods), gmm, and hierarchical clustering

For more advanced usage:

python3.6 main.py --season=<year> --num_clusters=<num> --interactive=<interactive_bool>

year needs to be of the form 20XX-XX+1; example: 2019-20. num from num_clusters needs to be an int, and specifies the max number of clusters the application will go to (from 3 to num). interactive_bool needs to be a bool, and specifies if you want to see an interactive plot on each clustering for each value of k clusters.

This repository now allows you to specify whatever year, and get all of the active players from that year.

Additionally, the img/ folder shows a bunch of pictures of clustering results among different fields. clustering methods, and values of k.

Due to some updates in the NBA stats API, there is the variable HEADERS defined in constants.py to ensure no connection times out.