What it does

AlphaPy is a Python framework that wraps scikit-learn, Keras, XGBoost, LightGBM, and CatBoost into two domain-specific pipelines: MarketFlow for financial time-series and SportFlow for sports prediction. It handles feature engineering, model training, ensemble stacking, and portfolio tear-sheet generation via Quantopian’s pyfolio. The goal is letting speculators run ML without writing boilerplate.

The interesting bit

The project explicitly splits its audience in two: “speculators and data scientists.” That honesty is refreshing. Most of the value is in the glue — standardizing a config-driven workflow so you can swap between Random Forest and a Keras net without rewriting data loaders. The sports-betting angle (SportFlow) is also unusual; most AutoML tools ignore that niche entirely.

Key highlights

• Supports blended and stacked ensembles out of the box
• Includes portfolio analysis via pyfolio integration
• Domain-specific pipelines: MarketFlow (trading) and SportFlow (sports)
• Legacy repo; active development moved to AlphaPy Pro (Python 3.12+, UV package management, NLP features, MetaLabeling)
• Pro version adds CI/CD, docs, and “enterprise-grade capabilities” (vague, but noted)

Caveats

• This repository is legacy; the README itself redirects users to AlphaPy Pro
• XGBoost requires manual platform-specific installation on Mac and Windows
• pyfolio dependency has a known numpy compatibility bug requiring a git install workaround
• No candidate images provided, and the README’s referenced image URLs appear to point to a different fork (Alpha314/AlphaPy)

Verdict

Worth a look if you want opinionated scaffolding for trading or sports ML and don’t mind migrating to the Pro fork for active support. Skip it if you need bleeding-edge research flexibility — this is automation for practitioners, not framework builders.