Abstract
This dissertation develops and evaluates a real-time adaptive robust portfolio optimisation (RTARPO) model designed to manage uncertainty across two contrasting asset classes, cryptocurrencies and equities. Using daily data from April 13, 2020, to August 31, 2025, the study examines how an adaptive robust framework performs against traditional strategies such as the equally weighted portfolio (EWP), mean-variance optimisation (MVO), and random optimisation (RO) under volatile and changing market conditions. The dataset, sourced from Yahoo Finance, captures major financial events including the COVID-19 recovery, the crypto boom and collapse of 2021-2022, and the inflationary tightening of 2022-2023. By integrating exponentially weighted moving average (EWMA) volatility estimation into ellipsoidal uncertainty models, the RTARPO approach adapts to changing risk dynamics in real time while protecting against estimation errors. The empirical findings reveal that the model consistently delivered stronger risk-adjusted returns and more stable Sharpe ratios than the benchmark strategies, particularly during periods of heightened market turbulence and in highly volatile environments such as the cryptocurrencies and equities markets, within the studied time horizon. Moreover, the mixed portfolio that combined cryptocurrencies with equities achieved the most effective balance between growth potential and stability, demonstrating the value of diversification across asset classes with differing risk characteristics within the studied time horizon. In calmer, more stable market environments, such as equities, traditional models like MVO tend to deliver stronger results, particularly when investors display low risk aversion. However, as risk aversion increases, random allocation strategies begin to perform better, while the RTARPO model steadily strengthens, demonstrating its ability to maintain resilience and stability when investors become more sensitive to downside risks. The study advances the theory of robust optimisation by linking it to real-time volatility modelling and it provides practical insights to portfolio managers on adjusting learning speeds and uncertainty parameters across different market regimes. From a policy perspective, the findings highlight the importance of incorporating adaptive, data-driven approaches into financial regulation and risk oversight to enhance resilience and stability in today’s increasingly digital and interconnected markets.