Rising Bandits for Best Athlete Identification
- Conducted a literature review on rising and rotting rested bandits (subset of multi-armed bandits), compiling use cases, regret bounds, and regret measures
- Leveraged successive rejection algorithms to estimate early career baseball players’ underlying OBP with confidence intervals
- Constructed a dataset of Retrosheet play-by-play data simulating Bernoulli rewards of plate appearance results
I’d like to revisit this at some point and see if this Bandits approach can be used to update projections faster in baseball