PyPortfolioOpt

PyPortfolioOpt makes the assumptions behind an allocation separate components. Expected returns describe the forecast; the covariance model describes how asset returns vary together. An EfficientFrontier object then combines those inputs with an objective and any added restrictions. You can supply your own estimates or calculate them from historical price data.

The example workflow loads prices, estimates returns and covariance, calls an optimizer, and cleans the resulting weights. save_weights_to_file exports them for later use. DiscreteAllocation takes those weights, recent prices and a total portfolio value to calculate whole-share quantities and leftover cash. It does not place trades.

Constraints can express a minimum portfolio score or a tracking-error limit against benchmark weights. A maximum number of holdings is more demanding: it needs a suitable mixed-integer solver, and the FAQ notes that this formulation does not work directly with every objective, including the standard maximum-Sharpe setup.

Use pip install PyPortfolioOpt in a Python environment and check the solver dependencies for the intended formulation. No PyPortfolioOpt login is required. The project distributes the library under the MIT licence; market data and any separately licensed solver remain their own inputs. Allocation estimates are sensitive to the return and risk assumptions, so changing an estimator can materially change the weights.