Interactive Firework Visualization of the October 10, 2025 Hyperliquid Liquidation Cascade
A real-time visualization of the largest documented DeFi liquidation event, showing 98,620 liquidations and auto-deleveraging (ADL) events as fireworks.
Live Demo: hyperfireworks.vercel.app
- CASCADE_TIMING_ANALYSIS.md added - First documentation of ADL activation delay
- 61-second delay discovered - 710 liquidations before first ADL
- 0.946 correlation proven - Liquidations directly cause ADL activation
- Burst pattern identified - ADL activates in waves, not continuously
- Visual validation - The "chunks" you see are REAL cascade mechanics!
- Fixed Double Counting: Stats now accurately progress in real-time without re-summing on replay
- Fixed Animation Freezing: Enhanced error handling prevents animations from freezing or not showing
- Fixed Progress Bar Stuck: Timeline and progress bar never get stuck randomly anymore
- Improved Data Accuracy: Events counted exactly once, maintaining data representation accuracy
- Better Performance: Particle limits (500 mobile, 1000 desktop) prevent performance degradation
- Enhanced Stability: Comprehensive try-catch blocks ensure continuous operation even if errors occur
processedEventIndicesSet tracks which events have been counted- Stats automatically recalculate when jumping backwards on timeline
- Animation loop always requests next frame first (prevents freeze)
- Invalid event and particle validation before processing
- Safety checks for elapsed time calculations
- Proper cleanup on reset and replay
Technical Analysis: See HyperMultiAssetedADL Repository
The visualization shows large chunks of red fireworks (liquidations) followed by green/red chunks (ADL). This is NOT a rendering artifact - it's revealing Hyperliquid's internal processing architecture!
| Discovery | Finding |
|---|---|
| First liquidation | Second 0 (event starts) |
| First ADL | 61 seconds later |
| Liquidations before ADL | 710 events |
| Correlation | 0.946 (liquidations predict ADL) |
| Biggest burst | 22,558 events/second (11,279 liqs + 11,279 ADLs) |
Pattern:
0-60s: Liquidations only - 710 events
↓ ADL threshold reached!
61s: MASSIVE burst - 22,558 events!
61-180s: Alternating waves of liquidations and ADL
Even within the same millisecond, events fire in separate batches!
At the biggest burst (21:16:04.831874), 22,558 events occurred, but they're NOT mixed:
Events 710 - 11,988: ALL 11,279 liquidations
Events 11,989 - 23,267: ALL 11,279 ADLs
Zero interleaving. Perfect separation.
The visualization reads events in their blockchain order. When you see:
- Red chunk - Hyperliquid's liquidation engine processing
- Green chunk - Hyperliquid's ADL engine processing
- Clear separation - Sequential batch execution, not concurrent
The chunks reveal Hyperliquid's internal architecture:
Each burst processes as:
├─ Phase 1: ALL liquidations (liquidation engine)
├─ Phase 2: Calculate ADL needs
└─ Phase 3: ALL ADLs (ADL engine)
Same timestamp, but ORDERED execution!
- ADL is NOT instantaneous - 61-second delay before activation
- ADL activates in BURSTS - threshold-based, not continuous
- Liquidations ALWAYS first - 100% of analyzed events show this order
- Sequential, not concurrent - Clear execution phases
- You can watch it unfold - The chunks you see are REAL processing batches!
- Macro: 61-second delay, burst patterns (CASCADE_TIMING_ANALYSIS.md)
- Micro: Sequential batch processing within each burst
- Visual: Your eyes see the actual protocol architecture in action!
This is the first visual demonstration of DEX internal processing architecture from blockchain data!
Full Technical Analysis: HyperMultiAssetedADL Repository
Batch Processing Details: BATCH_PROCESSING_DISCOVERY.md
Event: October 10, 2025, 21:15-21:27 UTC (12 minutes)
Platform: Hyperliquid L1 (Decentralized Perpetuals Exchange)
Total Market Impact: $7.61 BILLION
| Type | Events | Volume | Description |
|---|---|---|---|
| Liquidations | 63,637 | $5.51B | Users hit liquidation prices |
| ADL (Auto-Deleveraging) | 34,983 | $2.10B | Profitable positions force-closed |
| TOTAL | 98,620 | $7.61B | Complete cascade impact |
Full Analysis: ADL_PRIORITIZATION_VERIFIED.md
When you watch the green ADL fireworks, you're witnessing profitable traders being forced to exit. This is counterintuitive but proven by clearinghouse data:
| What People Think | Reality (Proven) |
|---|---|
| ADL targets high leverage | ADL targets high profit |
| Low leverage = Safe | High profit = Target |
From analyzing 31,444 ADL events with complete account data:
| Metric | Value |
|---|---|
| Profitable positions ADL'd | 98.3% |
| Average unrealized PNL | +82.43% |
| Average leverage | 1.16x (LOW!) |
Translation: Almost every green firework you see was someone with a large unrealized gain getting forcibly exited.
- Liquidations occur (red fireworks) - traders lose money
- Protocol needs counterparties to close these positions
- Protocol targets the MOST PROFITABLE opposing positions
- Forces them to exit (green fireworks) - winners pay for losers
Example from the visualization:
- $193M BTC position (the largest firework) = +12.73% profit, 6.3x leverage
- $174M ETH position = +21.84% profit, 5.6x leverage
- Both were ADL'd because of profit, not leverage
Green fireworks = Winners getting kicked out
Red ADL fireworks = Losers getting kicked out (rare - only 1.7%)
The cascade is forcing the most profitable traders to subsidize the liquidated traders.
- Green Fireworks: ADL with positive PNL (profitable position closed)
- Red Fireworks: ADL with negative PNL (losing position closed)
- Purple: Mega liquidations (> $10M)
- Red: Large liquidations ($1M - $10M)
- Orange: Medium liquidations ($100K - $1M)
- Gold: Small liquidations (< $100K)
Size varies logarithmically with the amount liquidated/ADL'd:
- Larger amounts = Bigger, more particles
- Smaller amounts = Smaller, fewer particles
| Control | Function |
|---|---|
| Auto-Start | Visualization begins automatically when page loads |
| Start | Begin/replay the visualization |
| Pause | Pause/resume playback |
| Reset | Reset to beginning |
| Speed | Cycle through playback speeds (1x Synced, 2x, 5x, 10x) |
| Timeline | Click/touch anywhere on the timeline to jump to that moment |
The visualization is synced with Tchaikovsky's 1812 Overture:
- The music starts automatically with the visualization (click anywhere if blocked by browser)
- The largest liquidation ($193M BTC) fires exactly at 15:28 during the most dramatic moment
- Music loops automatically when finished
- Synchronized timing for maximum impact
The interface displays real-time stats:
- Events Fired: Total fireworks launched
- Liquidations: Number of liquidation events
- ADL Events: Number of auto-deleveraging events
- Total Volume: Cumulative $ amount
- Timeline: Current position in the 12-minute event
Blockchain-Verified Data:
- Source: Hyperliquid S3 buckets (
node_fills_20251010_21.lz4) - All events explicitly labeled by the blockchain
- No heuristics or estimations
- Complete 12-minute dataset
Related Research:
- Full analysis: HyperMultiAssetedADL
- Position-level data: HyperAnalyzeADL
- Python 3.6+ (for data preparation)
- Modern web browser (Chrome, Firefox, Safari, Edge)
- Clone the repository:
git clone https://github.com/ConejoCapital/HyperFireworks.git
cd HyperFireworks- Prepare the data (if rebuilding):
python3 prepare_data.py- Start local server:
python3 -m http.server 8000 --directory public- Open in browser:
http://localhost:8000
This project is pre-configured for Vercel deployment:
- Push to GitHub
- Import repository in Vercel dashboard
- Deploy automatically
npm install -g vercel
vercel --prodvercel.json: Pre-configured for static site hostingpackage.json: Metadata and scripts
HyperFireworks/
├── public/
│ ├── index.html # Main HTML file
│ ├── fireworks.js # Animation logic
│ ├── events.json # Event data (98,620 events)
│ └── summary.json # Summary statistics
├── CASCADE_TIMING_ANALYSIS.md # Liquidation→ADL timing research
├── ADL_MECHANISM_RESEARCH.md # ADL counterparty analysis
├── prepare_data.py # Data preparation script
├── package.json # NPM configuration
├── vercel.json # Vercel deployment config
└── README.md # This file
- Particle-based firework explosions
- Gradient background
- Realistic physics (gravity, friction, decay)
- Color-coded by event type and PNL
- Size-coded by amount
- Real-time statistics dashboard
- Timeline with progress bar
- Interactive playback controls
- Event details on firework launch
- Legend for easy interpretation
- Efficient canvas rendering
- Music-synced playback (1x, 2x, 5x, 10x)
- Optimized particle system (50-70% reduction)
- Fully mobile-optimized with touch support
- 1812 Overture synchronized to largest event at 15:28
- Auto-starts on page load
- Clickable/touchable timeline navigation
When each firework launches, you'll see:
- Event Type: "LIQUIDATED" or "ADL"
- Ticker: Asset symbol (BTC, ETH, SOL, etc.)
- Amount: Notional value ($)
- PNL: Realized profit/loss
- Address: Wallet address (truncated)
This visualization captures the largest documented on-chain liquidation cascade in DeFi history:
- $7.6 BILLION in forced closures
- 98,620 events in 12 minutes
- 175 unique assets affected
- 8,218 events per minute sustained
- Protocol remained solvent
- ADL mechanism worked as designed
- No insurance fund bankruptcy
- Complete blockchain verification
- 1,000x larger than MakerDAO Black Thursday ($8M bad debt)
- First complete multi-asset DeFi cascade analysis
- Blockchain-verified with full transaction trails
This visualization demonstrates:
- Liquidation cascade mechanics in real-time
- ADL trigger patterns and distribution - See CASCADE_TIMING_ANALYSIS.md
- Market concentration effects (65% in top 3 assets)
- System resilience under extreme stress
- Event velocity and temporal patterns
- ADL activation delay - First documentation of 61-second threshold
- Burst-pattern behavior - ADL activates in waves, not continuously
- Liquidation→ADL causation - 0.946 correlation proves relationship
- Sequential batch processing - Liquidations → ADL (never concurrent)
- Internal architecture visibility - Visual chunks reveal protocol design
The visualization accidentally became a window into protocol internals! By displaying events in blockchain order, it reveals Hyperliquid's sequential batch processing architecture - liquidation engine runs first, ADL engine runs second, even when stamped with the same millisecond.
Event: October 10, 2025 Liquidation Cascade
Platform: Hyperliquid L1
Total Impact: $7.61 billion USD
Duration: 12 minutes (21:15-21:27 UTC)
Events: 98,620 forced closures
Assets: 175 unique tickers
Visualization: HyperFireworks (https://github.com/ConejoCapital/HyperFireworks)
- Frontend: Pure HTML5, CSS3, JavaScript (ES6+)
- Animation: HTML5 Canvas API
- Data Format: JSON (pre-processed from CSV)
- Hosting: Vercel (static site)
- Data Processing: Python 3 + Pandas
GPL-3.0 License - See LICENSE file for details
Created by: Conejo Capital
Data Source: Hyperliquid S3 Buckets
Research: HyperMultiAssetedADL
- HyperMultiAssetedADL - Full cascade analysis
- HyperAnalyzeADL - Position-level data
- HyperFireworks - This visualization
For questions, issues, or research collaboration:
- Open an issue on GitHub
- Related repositories: See links above
Special thanks to:
- Hyperliquid team for open S3 data access
- SonarX for data quality verification
- The DeFi research community
Watch the cascade unfold in real-time.
This visualization brings data to life, showing the largest DeFi liquidation event ever documented.
Live Demo: hyperfireworks.vercel.app