evaluating_nba_teams

Extending the concepts of adjusted plus/minus and complementary play styles to evaluate full NBA lineups and rosters.

Project Organization

Below is a brief summary of the main files and directories in the repo; each top-level entry is a file or directory in the root of the repo.

• data: contains all data that is written to disk, typically in CSV form
  • pbp: contains cleaned, scraped play-by-play data
  • testing: contains results from testing integrity of data
  • profiles: contains player profile data computed from pbp data
  • models: contains output from model fitting and performance evaluation
  • results: contains output from applications of the model
• src: contains Python source code
  • data: code for scraping and cleaning PBP data
  • testing: code for testing integrity fo PBP data
  • features: code for generating player profiles
  • models: code for training, selecting, comparing, and applying models
  • visualizations: code for generating visualizations and results
• slurm: contains slurm scripts for running code on Odyssey research compute cluster
• reports: home for all report-related source and output
  • fall_submission: report for fall CS91r
  • thesis: the actual thesis LaTeX and output
• models: contains pickled trained models that can be read from disk

---

Project based on the cookiecutter data science project template. #cookiecutterdatascience