Quantitative Finance / Independent exploration
Regime-Shift
Regime-conditioned portfolio research with a clear distinction between training-window isolation and causal test-time inference.
Problem
Market regimes can alter the relationship between risk and allocation. This project combines Gaussian HMM inference with regime statistics and portfolio weights.
Architecture
Feature preparation, HMM fitting, regime estimation, allocation, rebalancing, and reporting are separate modules. A rolling or expanding walk-forward validator refits the scaler and HMM for each training fold.
Implementation
Training-only scaler fitting is implemented. However, the entire test window is passed to Viterbi decoding and forward-backward posterior inference. The portfolio simulator then applies date-specific weights to that same date's returns.
Results
VERIFIED by source inspection: fold-specific fitting and overlap removal exist. The inspected inference/accounting path does not justify the report's leakage-free out-of-sample claim. Published return tables are not repeated here as validated outcomes.
Limitations
Future observations within a test sequence can affect smoothed state estimates. Same-day signal/return timing also needs correction or explicit justification. Failed folds are skipped after logging, so the successful-fold denominator matters. Fixed-universe and capacity assumptions remain research limitations.
Reproducibility
Reviewed at commit f1f4ea017b78 on 2026-11-09. Source inspection establishes implementation structure; repository-authored results are not independent measurements. The upstream application and benchmark suite were not rerun during this review.
Next investigation
Resolve the limitations above, preserve the reviewed commit and exact experiment configuration, and reproduce correctness checks before promoting performance or deployment claims.