Value-in-Vogue (VinV) is a downstream, regime-conditioned equity research system designed to evaluate capital behavior of dividend-persistent, value-oriented equities under fixed macro-financial regimes.
VinV is descriptive, not predictive.
It applies disciplined statistical & machine-learning methods only after macro regimes are defined upstream, translating observed equity behavior into auditable, reproducible research artifacts suitable for academic, institutional, & FinTech review.
VinV does not forecast returns, optimize portfolios, or generate allocation signals.
VinV operates within a strictly ordered research stack:
-
the Spine
Canonical macro-financial data fusion layer
(single source of truth, frequency normalization, locked semantics) -
FT-GMI
Regime-aware macro-financial diagnostics
(defines regimes & stress windows; read-only) -
VinV (this repository)
Equity behavior evaluation conditioned on FT-GMI regimes -
the OracleChambers
Interpretive narrative & validation layer only
(no signal creation, no feedback upstream)
Dependency is strictly one-way:
the_Spine → FT-GMI → VinV → OracleChambers
VinV never alters upstream data, regimes, or diagnostics.
- Evaluate equity behavior within fixed macro regimes
- Apply time-respecting walk-forward validation
- Translate probabilistic outputs into portfolio-level descriptive outcomes
- Produce frozen, audit-ready artifacts
- Prioritize stability, interpretability, & governance over model complexity
- Forecast returns
- Time markets
- Optimize allocations
- Generate trading signals
- Feed outcomes back into diagnostics
VinV uses machine-learning as an implementation tool, not an authority.
Modeling follows a deterministic ladder:
- Baselines → Linear → Tree → Boosted
- Leakage-safe preprocessing (X-only)
- Rolling features & winsorization
- Time-respecting walk-forward validation
- Empirical selection (no manual overrides)
- Portfolio-level evaluation (not point forecasts)
No deep learning.
No novelty chasing.
Stability & governance dominate complexity.
All VinV modeling derives from frozen, canonical parquet artifacts produced by the Spine.
the_Spine/vinv/ssot/
├── vinv_ml_ssot_vFinal.parquet
├── vinv_modeling_view_vFinal.parquet
- Champion: tier1_logistic_l2
- Selection basis: highest walk-forward stability
- Final refit cutoff: 2025-11-01
The champion is selected empirically, not heuristically.
This reflects maturity and governance discipline, not underfitting.
the_Spine/vinv/champion/vinv_champion_freeze_20251215T213325Z/
├── champion_model.joblib
├── champion_freeze_manifest.json
├── freeze_hashes_sha256.json
Controls enforced:
- Deterministic selection policy
- Timestamped freeze
- SHA-256 hash locking
- Git-tagged release
These artifacts constitute the primary empirical evidence for VinV.
VinV aligns with PMI-CPMAI principles:
- Simpler methods before complexity
- Separation of data, diagnostics, & interpretation
- Full provenance & versioning
- Anti-leakage controls
- Immutable evidence artifacts
This repository intentionally contains:
- Code
- Results evidence
- Governance artifacts
Raw licensed data is excluded by design.
Once FT-GMI is finalized, VinV can support:
- Cross-regime equity comparisons
- Stress-aware evaluation extensions
- Broader universes (e.g., WRDS-backed)
- Interpretive overlays via OracleChambers
All extensions preserve the same governance posture.
VinV is a bounded, governed, regime-conditioned equity research system that evaluates how equity constructions behave under defined macro contexts.
It is:
- Diagnostic, not predictive
- Applied, not promotional
- Stable, auditable, & review-ready
This repository is for research and educational purposes only. It does not constitute investment advice, forecasts, or recommendations.
MIT License