644 stars · Apache-2.0 · 0.4.5 (2025-07-23) · Track this in Rising ↗
Factor analysis for trading signals: does this feature actually predict returns, before a strategy is built on it.
Radar check
What we found when we opened it. Every bar is one fact, with its date.
Last code pushed 26 Oct 2026, one day before this check.
Latest release 0.5.1 on 26 Oct 2026.
Public since February 2021, a little over 5 years.
677 stars and 154 forks.
21 commits and 9 new issues in the last 90 days.
34 contributors. 21 pull requests merged in 90 days.
No security policy file says how to report a problem. No published security advisories.
How the bars are scored
- Alive. Days since the last code push: a week or less 100, a month 90, three months 70, six months 50, a year 30, eighteen months 15, older 0.
- Releases. Age of the latest GitHub release, on a similar scale; no bar when the project publishes no releases.
- Maturity. Age of the project: eight years 100, five 90, three 70, two 55, one 40, six months 25, younger 10.
- Adoption. Stars on a log scale: 100,000 stars is 100, 10,000 is 80, 1,000 is 60, 100 is 40.
- Activity. Commits in the last 90 days: 300 or more 100, 100 85, 30 65, 10 45, any 25.
- Community. People who have contributed: 100 or more 100, 30 80, 10 60, 5 40, 2 25, one 10.
- Security. A published security policy 80, none 40; published advisories are listed, not scored.
Checked by Repo Radar on 27 Oct 2026 from the repository on GitHub. Covered in Edition 17.
▶Repo detailsthe review · specs · pros & cons · install
What it does
It takes a factor, which is any number computed per instrument per day, together with the prices that followed. It reports the returns by quantile, the information coefficient, the turnover and the decay of the signal over different holding periods. The output is a set of charts that are the standard vocabulary of factor research, so the results are easy to discuss with anybody in the field. It is a maintained continuation of Quantopian's Alphalens, kept going by Stefan Jansen. Version 0.5.0 was published on 17 October 2026, the day of this refresh, and it needs Python 3.11 or newer. It analyses a signal and nothing more: it does not run a backtest, size positions, or place orders.
Why it matters
Who it suits. Anybody researching trading signals. It sits before the backtester, not after it. A backtest says what one set of rules did; this says whether the underlying signal carries information at all, which is the cheaper and more honest question to ask first. Skip it if there is no signal yet, because there is nothing for it to measure.
Verdict. The best twenty minutes available to anybody about to spend a month on a strategy, and the cheapest way to be told no. Two limits. It was built around daily stock data, so applying it to intraday data means reshaping the inputs by hand. The charts also assume some prior knowledge, so expect to read about quantile spreads before the output means much. stefan-jansen/zipline-reloaded is the backtester from the same maintainer, for the step after this one.
Summary (Key Points)from the README
Key features
- Returns analysis: Surfaces return statistics and plots for an alpha factor.
- Information Coefficient analysis: Produces IC statistics and plots to gauge the factor's predictive power.
- Turnover analysis: Reports how much the factor's signal changes over time.
- Grouped analysis: Breaks down factor performance by group, such as sector.
- Tear sheets: Creates a factor "tear sheet" from a signal and pricing data in two steps.
- Integration: Works with the Zipline backtesting library and Pyfolio for portfolio performance and risk analysis.
- Python support: The next release supports Python 3.11-3.15, with pandas 3 required on Python 3.15 and pandas 1.5 or 2.x on earlier supported versions.
- Example notebooks: Provides Jupyter examples showing how to read and use the tear sheet.
Ways to install
| Platform or method | What it needs | Effort |
|---|---|---|
| pip | Python 3.11-3.15 and the listed dependencies (matplotlib, numpy, pandas, scipy, seaborn, statsmodels) | Low |
| conda | Conda packages, which the README says were qualified on Linux, macOS and Windows | Low |
| Master branch (development code) | Installing from the Alphalens repository's master branch | Higher, since it is unreleased development code |
Important
- Python version: Python 3.10 and older are no longer supported, though previous release artifacts remain available.
- pandas pairing: NumPy 2.0 or later requires pandas 2.2.2 or later, so older pandas versions should be upgraded before using NumPy 2.
- Maintenance scope: Alphalens Reloaded only receives bug, security and compatibility fixes. New development focuses on ML4T Diagnostic, which the README recommends for new projects.
- Previous releases: The source changes do not alter previously published releases.
- Development code: The master branch is development code and may be less stable than a released version.
Written by an AI model from the project’s README; not part of the review, and it can be wrong.
- It answers the cheap question before the expensive one is attempted.
- Its charts are the standard vocabulary of factor research, so the output travels well.
- Apache-2.0, so there is no licence trap in commercial use.
- It was built around daily stock data. Intraday data has to be reshaped first.
- The output takes some learning. The charts assume familiarity with quantile spreads.
- It is a maintained fork of a finished project rather than a fast-moving one, so new features are rare.
source venv/bin/activate pip install alphalens-reloaded # Documentation: https://alphalens.ml4trading.io
