Regime Conditioning for Minimum Variance Allocation: Evidence from a five stock Indian equity basket
We evaluate whether market-state conditioning improves a long-only minimum-variance allocator on five Indian equities. Both methods use a 504-return rolling history, a 40 percent per-asset cap, delayed execution and proportional trading costs. A three-state Gaussian hidden Markov model estimates regimes from the equal-weight basket return. The 2025 evaluation comprises 249 daily returns. The updated HMM allocator achieves net annualized Sharpe 0.1491, compared with 0.1331 for the unconditional baseline. The paired difference is 0.0160, with a 95 percent moving-block bootstrap interval [-0.0105, 0.0417]. All 70 primary fits converged under the stated likelihood criterion. Higher traded notional and an interval spanning zero limit the evidence of benefit. The study provides a reproducible comparison with explicit costs and uncertainty. The updated engine was evaluated retrospectively on an already inspected historical window, which limits confirmatory interpretation.
Authors
- Paaras Kokate
- Sahil Belchada
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-17
- DOI
- https://doi.org/10.5281/zenodo.24306673
- Primary Topic
- Risk and Portfolio Optimization
- Type
- preprint