Market Regime-Based Multi-Asset Portfolio Optimization Using Nested Clustered Optimization for Tactical Asset Allocation

Supriadi and Bambang Siswoyo

Papers from arXiv.org

Abstract: We investigate the interaction between market regime detection and Nested Clustered Optimization (NCO) for tactical asset allocation across heterogeneous multi-asset portfolios. The central empirical question is whether conditioning the NCO pipeline, encompassing covariance estimation, hierarchical clustering, and weight allocation, on latent market states genuinely enhances risk-adjusted performance, or conversely, whether estimation uncertainty within regime-filtered sub-samples induces severe turnover and performance drag. We implement a causal rolling Gaussian Hidden Markov Model (HMM) coupled with an adaptive rebalancing mechanism that distinguishes regime shifts from portfolio weight drift. The framework is evaluated on a monthly Indonesian multi-asset universe comprising sovereign bond yields (SBN), the LQ45 equity index, spot gold (XAU/IDR), foreign exchange (CNY/IDR), and Bitcoin (BTC/IDR) across a 56-month out-of-sample period (January 2021 to August 2025). Rigorous walk-forward backtesting reveals that standard regime-agnostic NCO achieves remarkable capital preservation and risk-adjusted efficiency, delivering a 29.97% cumulative return (5.78% annualized), a maximum drawdown of only 2.18%, a Calmar ratio of 2.65, and low monthly turnover of 5.41%. In contrast, conditioning NCO on an unconstrained HMM in a small multi-asset universe amplifies parameter estimation noise and induces excessive rebalancing (47 events; 24.58% average monthly turnover), which erodes net cumulative return to 1.60% and increases drawdown to 9.99%. These findings demonstrate that without explicit transition persistence or asymmetric downside objectives, regime-switching frameworks can degrade rather than improve clustered portfolio stability, establishing crucial empirical boundaries for machine learning asset allocation.

Date: 2026-10, Revised 2026-10
References: Add references at CitEc
Citations:

Downloads: (external link)
https://arxiv.org/pdf/2610.27751 Latest version (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:arx:papers:2610.27751

Access Statistics for this paper

More papers in Papers from arXiv.org
Bibliographic data for series maintained by arXiv administrators ().