Research papers and replication¶

Author: Artur Sepp

The methods of OptimalPortfolios are described in the papers below. This page lists each paper once, with the citation used everywhere on the site, what it contributes to the package, and what the repository’s paper folders let a public checkout reproduce.

Software citation: CITATION.cff.

How the papers are used¶

Pages cite a paper by section and equation. A result from a paper is quoted with its study design and is never restated as a general performance claim. Exhibits on this site are either regenerated from code and data tracked in the repository or synthetic analogues drawn from a page’s own script; figures of the papers themselves are not reproduced here. Research that has not been published is neither cited nor displayed.

Papers¶

Robust optimization of strategic and tactical asset allocation¶

Sepp, A., Ossa, I. and Kastenholz, M. (2026). Robust Optimization of Strategic and Tactical Asset Allocation for Multi-Asset Portfolios. The Journal of Portfolio Management, 52(4), 86–120. Publisher page; DOI 10.3905/jpm.2025.1.806; author-shared copy.

The ROSAA framework, of which optimalportfolios is the reference implementation. It contributes:

the covariance estimated with a hierarchical-clustering group LASSO factor model, assembled as \(\Sigma = \beta \Sigma_F \beta^{\top} + D\) (see factor covariance with HCGL and FactorLasso);
strategic allocation by constrained risk budgeting (see risk budgeting);
tactical allocation as alpha over a tracking-error budget against the strategic benchmark (see tactical allocation).

The ROSAA case study reports its study design and results and runs the same configuration offline.

Optimal allocation to cryptocurrencies¶

Sepp, A. (2023). Optimal Allocation to Cryptocurrencies in Diversified Portfolios. Risk, October 2023. Risk; SSRN 4217841.

The paper compares four allocation methods for a diversified portfolio with a cryptocurrency: equal risk contributions, maximum diversification, maximum Sharpe ratio and CARA utility under a Gaussian mixture fitted to returns. Each is an objective of the package (see choosing an objective, mean-variance objectives and CARA utility under Gaussian mixtures); the manuscript source is tracked in the repository. The cryptocurrency case study reports its study design and results and runs the four methods offline on the tracked price panel.

Capital market assumptions from multi-asset tradable factors¶

Sepp, A., Hansen, E. and Kastenholz, M. (2026). Capital Market Assumptions and Strategic Asset Allocation Using Multi-Asset Tradable Factors. Working paper, SSRN 6785958.

The paper derives capital market assumptions from multi-asset tradable factors and uses them for strategic asset allocation. On this site it is cited only in its public SSRN version. The MATF-CMA case study builds the workflow from CMAs to a strategic allocation with the package on synthetic inputs; it reproduces none of the paper’s results.

Reproducing the papers¶

The paper folders are repository-only research code: they are not installed by pip install optimalportfolios, and each folder’s README states its own commands and data requirements. The table summarises what a public checkout can run.

Paper	Folder	What a public checkout reproduces
ROSAA	`robust_optimisation_jpm_2026`	A methodological example of the HCGL covariance and risk-budgeted strategic allocation. It downloads its ETF panel with `yfinance`, carries no frozen inputs and records no environment, so it is not an exact rebuild of the published exhibits.
Cryptocurrencies	`crypto_allocation_risk_2023`	The manuscript source, the analysis code and historical price files; no CI workflow runs this folder’s tests. The cryptocurrency case study runs the four methods offline on the tracked price panel. The full update route needs licensed Bloomberg data, so the headline numbers are not promised to reproduce exactly.
Capital market assumptions	`cma_data`	A manifest-verified snapshot of the configuration tables behind the capital market assumptions, with tests that the required checks run on every pull request to `main`. Licensed index, factor-history and provider panels are omitted.

Frozen package versions are quoted only from a committed manifest; where a folder records no environment, none is inferred.

How to cite¶

Cite a paper for its method, and the software records for the packages an implementation uses: optimalportfolios, qis and FactorLasso. FactorLasso’s own papers are listed on its research papers page.