An enterprise-grade quantitative econometric system and machine learning framework designed to detect financial asset bubbles, diagnose macro regime shifts, and predict structural market crash probabilities.
🚀 Access the Live Panel WebAssembly Dashboard on GitHub Pages
(Compiled directly to WebAssembly using HoloViz Panel & Pyodide — runs 100% client-side in your browser with zero backend server required!)
The system supports dual calibration horizons for model parameter tuning and regime walk-forward validation:
| Horizon Option | Date Bounds | Regimes | Native Feature Fidelity | Target Coverage |
|---|---|---|---|---|
| Option 1: Modern 5-Regime Horizon | 2015-01-01 to 2026-07-28 |
5 Regimes | 100% Native High-Fidelity | 2018 Volmageddon, 2020 COVID Flash Crash, 2020–2021 Post-COVID Exuberance, 2022 Fed Rate Hikes, 2024–2026 AI CapEx Rally |
| Option 2: Expanded 7-Regime Horizon | 1998-01-01 to 2026-07-28 |
7 Regimes | ~92% Extended Historical Spectrum | Spans all 7 major crashes including 1999–2000 Dot-Com Peak (CAPE 44.19) and 2007–2009 Subprime GFC (Housing PTI ~7.0x) |
- Macro Valuation Anchors: Tracks Shiller CAPE (41.37), Payout-Adjusted CAPE (P-CAPE), and the Buffett Indicator (218.1% of GDP) against historical quintile bands.
- Systemic Liquidity & Leverage: Monitors FINRA Margin Debt ($1.416T Peak), YoY velocity, and excess debt capacity ("Margin Exhaustion Score").
- Econometric Explosive Bubble Diagnostics: Computes Generalized Supremum ADF (GSADF / PSY Procedure) with General-Purpose Technology (GPT) decomposition filtering $754B AI CapEx shocks.
- Options Sentiment & Behavioral Tracking: Analyzes VIX contango term structure (VIX1D vs VIX3M), CBOE SKEW tail-risk alerts (>145), and OVX/VIX cross-asset volatility decoupling.
- Sector Vulnerability & Topological Dynamics: Measures Housing Price-to-Income affordability (7.11x) alongside Topological Data Analysis (TDA) persistence landscape L2 norm and Morlet wavelet complexity scores.
git clone https.github.com/danieldf/bubble-detector.git
cd bubble-detector
uv venv
source .venv/bin/activate
uv pip install -r requirements.txtpython -m bubble_detector.ui.dashboardNavigates to http://localhost:8080
panel serve bubble_detector/ui/panel_dashboard.py --showNavigates to http://localhost:5006
panel convert bubble_detector/ui/panel_dashboard.py --to pyodide-worker --out dist/
python -m http.server 8000 --directory dist/pytest tests/MIT License. Developed for quantitative systemic risk assessment and econometric research.