Machine Learning Portfolio Modeling for Bitcoin Investment

This technical report discusses the application of machine learning portfolio modeling techniques to evaluate the impact of bitcoin on multi-asset class portfolios. It outlines a new modeling framework developed by Fidelity that addresses the limitations of traditional statistical models, particularly in the context of bitcoin's short price history and high volatility. The report presents a comparative analysis of portfolio performance with and without bitcoin, emphasizing the importance of personalized return assumptions and risk tolerance. It details how the model can simulate various market scenarios to provide probabilistic outcomes, allowing investors to understand the potential risk-reward profile of including bitcoin in their portfolios. Key observations indicate that bitcoin's return patterns resemble those of other emerging assets and that its inclusion may enhance a portfolio's risk-reward profile, depending on specific allocation constraints and expected return assumptions. The report concludes with a discussion on the need for ongoing model calibration to adapt to evolving market conditions.