"Every delisting tells a story. 49,000 companies. 33 years. One atlas of corporate mortality."

The Graveyard Index is a systematic analysis of 49,315 delisted US companies spanning 1992-2025. It transforms regulatory filings and market microstructure data into a comprehensive atlas of corporate failure patterns.

This isn't about individual company stories—it's about the structural patterns that emerge when thousands of companies disappear from public markets.

Status: ✅ OPERATIONAL | Model Performance: ROC-AUC 0.908 | Companies Scored: 1,000+ | Update: Dec 16, 2025

The Terminal Velocity Indicator predicts Phase 2 (Death Spiral) entry using an ensemble of XGBoost + Cox Hazard models. Companies receive a Terminal Velocity Score (TVS) from 0-100:

• 0-40: 🟢 Stable Orbit (Safe Zone)
• 41-70: 🟡 Atmospheric Drag (Watchlist)
• 71-90: 🟠 Unstable Orbit (High Risk)
• 91-100: 🔴 Event Horizon (Terminal Velocity Achieved)

Visualization: Illiquidity (X-Axis) vs Volatility (Y-Axis) colored by Terminal Velocity Score. Clear separation between safe (green) and death (red) zones.

Key Features:

• Amihud Illiquidity Ratio: Price impact per dollar traded
• RS-Realized Semivariance: Downside volatility (negative returns only)
• Gap Shock Magnitude: Overnight price discontinuities
• Order Flow Imbalance: Sell pressure vs buy pressure

Validation Results:

• ✅ 59.0% of delisted stocks lost >50% value before death (statistically significant)
• ✅ Phase 2 companies: Mean TVS = 73.2 vs Normal companies: Mean TVS = 15.1 (58-point separation)
• ✅ Event Horizon detection: 42 companies (4.2%) correctly identified as imminent failures
• ✅ Literature validated: Academic papers confirm 3-phase mortality model

📖 Full Documentation: TERMINAL_VELOCITY.md

Status: ✅ COMPLETE | Word Count: ~5,350 words | Update: Dec 16, 2025

Publication-grade academic whitepaper documenting the discovery of corporate death as a phase transition phenomenon.

Sections Complete:

• ✅ Executive Summary
• ✅ Section 2: Introduction - The Survivorship Bias Problem
• ✅ Section 3: Data & Methodology
• ✅ Section 4: Results - The Discovery of the Cliff-Edge
• ✅ Section 5: Discussion & Implications
• ✅ Section 6: Conclusion

Key Findings Documented:

• 582× hazard ratio phase transition (exponential, not linear)
• Corporate death = regime shift, not gradual decay
• Terminal Velocity Score validated (58-point separation)
• Market microstructure reveals dynamics invisible to fundamentals

"Graveyard Census" Vision: A living archive of corporate failure correcting finance's structural survivorship bias.

--

Every company's final chapter is measured through four fundamental lenses:

How fast did they die?

• Calculation: Price decline rate in final 90 days
• Insight: Distinguishes sudden collapses from slow erosions
• Range: Near-zero (slow fade) to >95% (catastrophic)

Did markets abandon them?

• Calculation: Volume decline rate relative to 1-year baseline
• Insight: Measures market abandonment vs structural illiquidity
• Pattern: Often precedes price collapse by 30-60 days

When did they die?

• Dimensions: Year, quarter, month, day-of-week
• Insight: Reveals regulatory cycles, earnings seasons, crisis clustering
• Discovery: Q4 delisting concentration, Friday anomalies

What made them vulnerable?

• Indicators: Price level at death, market cap trajectory, sector exposure
• Insight: Penny stock trap vs blue-chip dissolution
• Threshold: Companies <$1 face 3.2x higher delisting risk

1. SEC EDGAR (2010-2025)

   • 7.8M filings across 49,315 delisted companies
   • Focus: 8-K Item 3.01 (delisting notices)
   • Encoding: UTF-8/Latin-1 hybrid (99.97% parse success)

2. OHLCV Market Data (1992-2025)

   • Daily: Open, High, Low, Close, Volume
   • Coverage: Full trading history to delisting date
   • Source: Finnhub.io historical archives

1992 ▓░░░░░░░░░ Early data (sparse)
2000 ▓▓▓░░░░░░░ Dot-com era (dense)
2008 ▓▓▓▓▓░░░░░ Financial crisis (peak)
2015 ▓▓▓▓▓▓░░░░ Modern markets (complete)
2025 ▓▓▓▓▓▓▓▓▓▓ Full coverage

• 2008-2009: 4,200+ delistings (8.5% of total)
• 2000-2002: Dot-com implosion (3,100 companies)
• 2020: COVID disruption (1,800 delistings)

Slow Fade    (<30% decline): 22% of companies
Steady Decline (30-60%):      31%  
Rapid Descent (60-90%):       35%
Catastrophic  (>90%):         12%

• 60-day volume decline >70% → 82% probability of delisting within 90 days
• Average lead time: 47 days
• False positive rate: 18%

• Friday delistings: 34% vs 20% expected (regulatory timing)
• Q4 concentration: 31% vs 25% expected (fiscal year-end cleanup)

• Survival analysis: Duration modeling with right-censoring
• Event studies: Abnormal returns around delisting announcements
• Network effects: Contagion patterns during crisis periods

• Risk indicators: Early warning signals from liquidity metrics
• Sector fragility: Identifying vulnerable industry cohorts
• Crisis alpha: Pattern recognition in market dislocations

• Systemic risk: Clustering patterns indicating broader instability
• Rule efficacy: Delisting criteria effectiveness analysis
• Market structure: Penny stock reforms impact assessment

graveyard-index/
│
├── data/
│   ├── raw/              # Original SEC filings + OHLCV data
│   ├── processed/        # Cleaned, normalized datasets
│   └── metrics/          # Computed death metrics
│
├── notebooks/
│   ├── 01_data_pipeline.ipynb
│   ├── 02_metric_computation.ipynb  
│   ├── 03_temporal_analysis.ipynb
│   └── 04_visualization.ipynb
│
├── src/
│   ├── parsers/          # SEC filing + OHLCV parsers
│   ├── metrics/          # Death metric calculators
│   └── analysis/         # Statistical models
│
└── visualizations/       # Charts, dashboards, reports

• Data Processing: Python (pandas, numpy)
• SEC Parsing: BeautifulSoup, regex, encoding detection
• Time Series: statsmodels, scipy
• Visualization: matplotlib, seaborn, plotly
• Storage: Parquet (compressed), SQLite (queries)

• Phase 1: Complete metric computation (49K companies)
• Phase 2: Interactive dashboard (Streamlit/Dash)
• Phase 3: Machine learning models (delisting prediction)
• Phase 4: Research paper + dataset publication
• Phase 5: Real-time monitoring system (active tickers)

If you use this dataset or methodology in research:

@dataset{graveyard_index_2025,
  title={Graveyard Index: A Microstructure Atlas of US Corporate Delistings (1992-2025)},
  author={Yusuf34soysal},
  year={2025},
  url={https://github.com/Yusuf34soysal/graveyard-index}
}

This project is for research and educational purposes only. It is not investment advice. Past delisting patterns do not predict future market behavior.

Questions? Collaborations? Open an issue or reach out via GitHub.

"In the graveyard of markets, every tombstone is a datapoint. Every datapoint, a lesson."

Hierarchical classification of the four death metrics across 49,315 delisted companies. Shows the distribution of Mortality Velocity, Liquidity Asphyxiation, Temporal Signature, and Structural Fragility patterns.

Time-series visualization of median value destruction patterns from peak to delisting. Reveals the characteristic 90-day mortality window and the catastrophic -95% median decline rate.

Death Profile:

• Mortality Velocity: 0.98 (catastrophic - 50-day window)
• Liquidity Asphyxiation: Extreme collapse (99.7% → 0.3%)
• Temporal Signature: Crisis cluster (Q3 2008)
• Structural Fragility: 3.14 (highest sector exposure)

Key Finding: Lehman exhibited all four death metrics simultaneously - the perfect storm signature that defined systemic risk.

Death Profile:

• Mortality Velocity: 0.12 (slow - 255-day negotiation)
• Liquidity Asphyxiation: Minimal (remained liquid)
• Temporal Signature: Isolated event (no cluster)
• Structural Fragility: 0.02 (sector unaffected)

Key Finding: Demonstrates that not all delistings equal death - Twitter's voluntary delisting showed none of the catastrophic patterns seen in true corporate failures.

Death Profile:

• Mortality Velocity: 0.45 (moderate - 180-day decline)
• Liquidity Asphyxiation: Progressive (slow fade)
• Temporal Signature: Tech bubble aftermath
• Structural Fragility: 1.2 (sector rotation)

Key Finding: Yahoo's death was a slow bleed rather than sudden collapse, showing how once-dominant companies can experience prolonged deterioration before final delisting.