Dynamic Bayesian Black-Litterman model: Portfolio optimization and comparative analysis

Download This Article

This work is licensed under a Creative Commons Attribution 4.0 International License.

Abstract

Portfolio allocation in rapidly changing markets requires frameworks that can adapt to time-varying expected returns and covariance structures. This study develops and evaluates a dynamic Bayesian Black–Litterman (DBBL) model that extends the traditional Black–Litterman framework through recursive Bayesian updating, dynamic covariance estimation, and LSTM-generated return views. Using 11 U.S.-listed assets (AAPL, MSFT, AMZN, GOOG, TSLA, JNJ, JPM, NVDA, META, XOM, and GLD) and daily data from 2015 to 2025, the DBBL model is compared with Markowitz and static Black–Litterman benchmarks. The results show that DBBL provides a modest improvement in Sharpe Ratio, indicating better risk-adjusted efficiency. However, this gain is accompanied by lower cumulative returns and deeper maximum drawdown relative to the comparator models, highlighting a clear trade-off between adaptive risk management and absolute portfolio performance. These findings suggest that DBBL should be interpreted as a risk-aware adaptive allocation framework rather than a uniformly superior portfolio solution, and they underscore the importance of balancing responsiveness and stability in dynamic portfolio construction.

Keywords: Dynamic Bayesian Black-Litterman, Markowitz, Portfolio Optimization, Bayesian Inference, Sharpe Ratio, Dynamic Allocation, U.S. Equities, GLD

Authors’ individual contribution: Conceptualization—P.W.; Methodology—P.W.; Validation—T.C.; Formal Analysis—T.C. and T.K.; Investigation—P.W.; Resources—P.W.; Data Curation—P.W.; Writing—Original Draft—P.W.; Writing—Review & Editing—T.C. and T.K.; Visualization—T.C. and T.K.; Supervision—T.C. and T.K.; Funding Acquisition—P.W.

Declaration of conflicting interests: The Authors declare that there is no conflict of interest.

JEL Classification: C11, G11, G17

Received: 29.10.2025
Revised: 09.03.2026; 20.03.2026
Accepted: 20.04.2026
Published online: 24.04.2026

How to cite this paper: Wattanasin, P., Chomtohsuwan, T., & Kraiwanit, T. (2026). Dynamic Bayesian Black-Litterman model: Portfolio optimization and comparative analysis. Risk Governance and Control: Financial Markets & Institutions, 16(2), 8–15. https://doi.org/10.22495/rgcv16i2p1