Modeling Subset Behavior: Prescriptive Analytics for Professional Basketball Data

/python

• nearest_neighbor_lineups.py: Implements graphical model for predicting lineup plus-minus from individual player stats.
• stint.py: Implements the stint class used in nearest_neighbor_lineups.py.

/R

• create_stints.R: Parses play-by-play data into stint tables. Generates files needed to run other R scripts. Run this before other R files
• plot_*: Generates figures used in thesis report.
• scrapeIndividualStats.R: Code used for scraping basketball-reference.com. Should be used for reference only -- does not run as is.
• train_k_means_3.R: Runs k-means clustering on individual player data. Generates figures used in thesis report.
• create_interaction_RAPM_design_matrix.R: Adds interaction columns to stint table.
• train_test_split_stints.R: Makes train/test split for stint table.
• standardize_individuals_2015_2016.R: Generates standardized_aggregate_individuals_2015_2016.csv from aggregate_individuals_2015_2016.csv.
• shiny/*: Shiny app for visualizing individual player data.

How to Run RAPM Model

• If not already run, run create_stints.R then train_test_split_stints.R.
• Run regular_RAPM.R

How to Run Interaction RAPM Model

• If not already run, run create_stints.R then train_test_split_stints.R then create_interaction_RAPM_design_matrix.R.
• Run interaction_RAPM.R.

NBA Analytics Summer 2018

• create_stints2017_2018_mod.R: Parses play-by-play data into stint tables with counts for lineup statistics. Generates files needed to run other R scripts. Run this before other R files
• stint_parser: Parses and aggregates stint data into lineup data. Also parses and aggregates data from basketball-reference.com Run this right after create_stints2017_2018_mod.R
• johnson_graph.R: Generates adjacency graph of lineups by player differential.
• cluster_id_model.R: Gives the functions needed to do the Cluster+Lineup ID model descriped in paper.
• individual_lineup_regression.R: Uses the basic lm function and explores the relationships between individual players statistics, clusters, and teams.
• tda_cluster_lineups.R: Uses topological data analysis to cluster the lineups into lineup types.
• tda_cluster_individuals.R: Uses topological data analysis to cluster the individual players into new positions.
• scraper_player_stats.R: Scrapes individual player stats from the 2017-2018 season from basketballreference.com.
• ridge_lasso_regression_lineups.R: Runs the ridge regression and lasso regression model on the lineup data to try and predict plus minus.
• ridge_lasso_regression_individual_stats.R: Runs the ridge regression and lasso regression model on the binded individual player data to try and predict plus minus.
• logit_regression_indiv_stats.R: Runs logistic regression on the binded individual player data to try and predict whether a lineup will be positive or negative.
• knn_lineups_w_clusters.R: Implements KNN on the lineup data to verify the cluster that it was placed into.
• create_warriors_indiv_stats_df.R: Creates the binded individual player data frame from all combinations of the Golden State Warriors 2018-2019 roster.
• create_lakers_indiv_stats_df.R: Creates the binded individual player data frame from all combinations of the LA Lakers 2018-2019 roster.
• create_all_star_df.R: Creates the binded individual player data frame from all combinations of an all star team.