A white box, on purpose

1. Multi-Window Instability Engine

The MVIE composite score blends three rolling Z-scored metrics—Absorption Ratio (PCA variance concentration), Instability Score (correlation-structure change rate), and Priority Drift (MST centrality rank drift)—across 30/60/90-day windows:

# Pseudocode — MVIE composite pipeline# AR, IS, and PD are each rolling Z-score outputs (30/60/90-day windows)AR = z_score(pca_variance_share(asset_returns, top_components=3))IS = z_score(correlation_structure_change_rate(correlation_matrix, baseline))PD = z_score(network_centrality_drift(min_spanning_tree, window=rolling_window))MVIE = (0.40 * AR) + (0.35 * IS) + (0.25 * PD)

2. Network Topology Analysis

Spearman correlation builds a Minimum Spanning Tree (d_ij = sqrt(2(1 − ρ_ij))). Eigenvector, betweenness, and degree centrality on the MST quantify systemic hubs—and feed Priority Drift when rank order changes across rolling windows:

• Eigenvector centrality — Hub assets whose stress propagates through the MST (e.g., SPY)
• Betweenness centrality — Bridge nodes linking otherwise disconnected clusters (contagion paths)
• Degree centrality — Direct connectivity count; used with rank drift for Priority Drift (PD)

3. Probabilistic Strategy Engine

Composite regime readings (AR · IS · PD) feed a probabilistic scenario layer—illustrative BULLISH / BEARISH / NEUTRAL labels with volatility envelopes, aligned to the same macro-risk stack as the dashboard:

Composite weights (engine — matches MVIE pipeline):  Absorption Ratio (AR):           40%  Instability Score (IS):          35%  Priority Drift (PD):             25%Output:  Direction:    BULLISH | BEARISH | NEUTRAL  Probability:  P(direction) ∈ [0, 1]  Envelope:     ±2σ volatility band (illustrative)  Horizon:      5 trading days, best / base / worst projections

4. Veto Protocol

The Veto Protocol is a model override rule. When MVIE > 75 (RED regime), all BULLISH scenario labels are reclassified to NEUTRAL or defensive overlays in the published set, regardless of micro-indicator strength. In crisis regimes, the engine treats micro readings as high-noise; the defensive scenario narrative takes precedence in the illustrative output.

5. LLM Analyst Layer

A large language model layer synthesizes the quantitative output into a structured institutional report. The prompt is compliance-locked — no investment advice, only analytical scenarios. Output is markdown-formatted with explicit risk framework sections. The underlying model can be swapped as newer APIs become available.