Diamond Edge: MLB pick recommendation system — gradient-boosted models + grounded LLM rationale.

Overview

Two parallel Claude agent pipelines feeding a gradient-boosted classifier. Walk-forward backtest framework (mlb-market-models sister repo) with Bayesian shrinkage and isotonic calibration. Portfolio piece — free + informational.

Case study

Full write-up landing in a later milestone. For now, the canonical source is the GitHub repo — it contains the README, eval results, architecture notes, and current state of the project.