robertmartin8/PyPortfolioOpt

PyPortfolioOpt is a comprehensive portfolio optimization library for Python that provides a full suite of methods for constructing and analyzing investment portfolios. At its core, the library implements mean-variance optimization, the Black-Litterman Bayesian model, and Hierarchical Risk Parity, giving users multiple approaches to asset allocation. It includes a complete covariance estimation toolkit with interchangeable estimators such as sample, exponential, shrinkage, and minimum-covariance-determinant methods, along with expected return estimation using historical mean, exponential weighting, or the Capital Asset Pricing Model.

The library distinguishes itself through several advanced capabilities. It features a Black-Litterman framework that blends market-cap-weighted prior returns with user-specified views using Bayesian updating, complete with view confidence specification and posterior covariance computation. The efficient frontier is computed using a dedicated Critical Line Algorithm solver that handles inequality constraints natively and converges in a fixed number of steps. For practical implementation, the library includes a discrete allocation rounder that converts continuous portfolio weights into integer share counts while respecting minimum trade sizes and cash balance. Hierarchical Risk Parity allocation builds diversified portfolios by recursively bisecting the correlation matrix with single-linkage clustering and allocating variance across clusters without inverting the covariance matrix.

The optimization engine supports a wide range of constraints including bounds on position size, long/short limits, and market neutrality, and can optimize for alternative risk measures such as semivariance and Conditional Value-at-Risk. Users can also integrate custom optimization algorithms by inheriting from base optimizer classes while reusing built-in weight-cleaning and performance-reporting utilities. The library provides tools for deriving market-implied prior returns from market-cap weights, computing posterior return weights, and maximizing the Sharpe ratio or minimizing volatility through the Critical Line Algorithm.