Oliveira–Newell (ON) score pipeline
May 1, 2024
·
1 min read
This repository implements the methodology from “Unlocking Athlete Performance: A multilevel approach for evaluating contributions in basketball.” It combines mixed-effects models and PCA to generate the Oliveira–Newell (ON) score at season, game, and athlete-contribution levels.
Components
- Data wrangling scripts for NBA play-by-play/box-score data (2015–2019).
- Mixed-effects model fitting that respects the hierarchical (player–team–season) structure.
- PCA-based aggregation of model outputs into a single ON score.
- Validation scripts comparing predictions across seasons/games.
Why it matters
- Provides a calibrated, reproducible alternative to ad-hoc box-score metrics.
- Fast (restricted likelihood) estimation compared to Bayesian samplers; suitable for season-long dashboards.
- Extensible to other sports with similar hierarchical data structures.
Run the scripts per the README to fit models, compute ON scores, and reproduce the analyses demonstrated in the paper.
