AlphaFactory-Pro

This repo is a reproducible, end-to-end “alpha research loop” for market-neutral, medium-frequency equity signals:

1. download/cached data (default: Yahoo via yfinance)
2. compute factor library (signals)
3. run purged walk-forward validation
4. build daily market-neutral long/short portfolios
5. apply turnover-aware transaction costs
6. combine signals using an online L1-budget allocator (differentiator)
7. generate a report (CSV + Markdown + plots)

⚠️ Research/education only. Not investment advice.

---

Motivation

If we only “try random alphas” on platforms (WorldQuant BRAIN etc.), it’s easy to get stuck. This repo forces a real research workflow: hypothesis → signal → strict evaluation → robustness → library → combination.

---

Quickstart

python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
pip install -e .

# run end-to-end pipeline (downloads/caches data, then backtests)
python -m alphafactory.run --config configs/base.yaml

Outputs go to: results/<timestamp>/

• factor_ic_summary.csv — factor IC / ICIR by split
• portfolio_daily_returns.csv — daily L/S returns (gross + net of costs)
• report.md — summary + key plots

---

Data notes

Default data is downloaded with yfinance for convenience. This is not survivorship-bias free and is not “publication-quality”. For production-grade research we can use CRSP/Compustat (WRDS), Norgate, etc.

This repo is built so you can swap in a different data source later:

• see src/alphafactory/data/

---

Repo structure

alpha-factory-pro/
  configs/
    base.yaml
  src/alphafactory/
    data/
    features/
    validation/
    portfolio/
    allocator/
    metrics/
    reports/
    run.py
  tests/
  paper/one_pager.md
  results/   # gitignored

---

References

• Qlib (data + ML quant pipeline): https://github.com/microsoft/qlib
• Alphalens (factor analysis): https://github.com/quantopian/alphalens
• vectorbt (fast vectorized research/backtests): https://vectorbt.dev/
• Ken French data library (FF factors for attribution): https://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html