fincore | Quantitative Performance & Risk Analytics

Documentation · 0.5 migration · Contributing · Changelog

What fincore 0.5 is

fincore is a unified Python platform for quantitative performance analysis. It keeps the analytical capabilities historically associated with Empyrical, Pyfolio, and Alphalens, but rebuilds them as one low-coupling core rather than three package-shaped APIs. The public contract is organized by domain and each capability has one canonical implementation path.

Version 0.5.1 is a quality-maintenance release that preserves the deliberately breaking 0.5.0 public contract. It does not provide fincore.empyrical, fincore.pyfolio, fincore.alphalens, flat root metric functions, compatibility aliases, or façade classes. Update imports to the focused domain modules described in the migration guide.

The package root is only a namespace index. Import executable functions and models from their owning leaf module, not from fincore itself.

Install

pip install fincore

# Optional capabilities
pip install "fincore[factor-analysis]"
pip install "fincore[visualization]"
pip install "fincore[interactive]"
pip install "fincore[report-pdf]"
pip install "fincore[report-xlsx]"
pip install "fincore[bayesian]"
pip install "fincore[data-yahoo]"
pip install "fincore[data-alphavantage]"
pip install "fincore[data-pandas-datareader]"
pip install "fincore[data-cn]"
pip install "fincore[all]"

For a source checkout, use pip install -e ".[dev]". Python 3.11+ is required. pyproject.toml is the dependency source of truth.

Quick start: metrics

import pandas as pd

from fincore.metrics.drawdown import max_drawdown
from fincore.metrics.ratios import sharpe_ratio
from fincore.metrics.yearly import annual_return

returns = pd.Series([0.01, -0.005, 0.002, 0.004])

print(sharpe_ratio(returns))
print(max_drawdown(returns))
print(annual_return(returns))

Portfolio report workflow

Compute a portable report model once, then choose a renderer. Renderers do not repeat analytical computation.

import pandas as pd

from fincore.report.portfolio.compute import build_portfolio_report
from fincore.report.renderers.html import write_html

index = pd.date_range("2024-01-02", periods=5, freq="B")
returns = pd.Series([0.01, -0.005, 0.002, 0.004, -0.001], index=index)
positions = pd.DataFrame({"AAA": 100.0, "BBB": -30.0, "cash": 80.0}, index=index)

document = build_portfolio_report(returns, positions=positions, rolling_window=3)
artifacts = write_html(document, "portfolio-report.html")
print(artifacts.named_artifacts["file"])

Use fincore[report-pdf] or fincore[report-xlsx] only when selecting the corresponding renderer.

Factor analysis

Factor analysis is a first-class domain, not an Alphalens wrapper. The repository includes an offline deterministic quickstart that prepares inputs, computes the canonical model, derives portfolio inputs, and renders a headless summary:

pip install "fincore[visualization]"
MPLBACKEND=Agg python examples/factor_analysis_quickstart.py

For application code, import the exact operation from its owning module, such as fincore.factor_analysis.data, analysis, performance, portfolio, or inference.

Design commitments

• One implementation path per public operation; no re-exported legacy API shells.
• Immutable input and extension snapshots at domain boundaries.
• Structured runtime errors with operation and parameter context.
• Report models separated from renderers and artifact lifecycle management.
• Optional dependencies isolated by explicit capability extras.
• Tests protect canonical imports, removed legacy surfaces, report semantics, package contents, and executable documentation.

Development and verification

/Users/yunjinqi/opt/anaconda3/bin/conda run -n base python -m pytest -o addopts='' tests/docs tests/packaging -q
/Users/yunjinqi/opt/anaconda3/bin/conda run -n base python -m ruff check fincore scripts tests

The release-quality and provenance records live under docs/quality/. A local test pass is not a published-release claim.

License and third-party notices

fincore is MIT licensed; see LICENSE. Retained third-party material keeps its own licensing and attribution in NOTICE, THIRD_PARTY_NOTICES.md, and THIRD_PARTY_LICENSES.