VaR requires stable mean, variance, and covariance. In crypto, all inputs are unreliable and change constantly. VaR is meaningless given fat tails.
Hypothesis HY10034
VaR requires stable mean, variance, and covariance. In crypto, all inputs are unreliable and change constantly. VaR is meaningless given fat tails.
Trading hypothesis
What traders get wrong
False assumption:
"VaR based on mean, variance, covariance is useful."
Truth:
All VaR inputs are unreliable in crypto. Fat tails make normal-distribution VaR meaningless.
Problem for trader:
Mean shifts constantly. Variance explodes without warning. Correlations spike during stress.
Key takeaways
What you should consider as a trader
- Mean is unstable - Expected return shifts dramatically.
- Variance explodes - Can 10x overnight.
- Correlations spike - Assets that seem uncorrelated become correlated.
- Fat tails break VaR - 99% VaR gets breached 5% of the time.
- Recommended: use multipliers - If using VaR, multiply by 2-3x.
Data you need
Assess VaR reliability
Data points:
- VaR breach frequency
- Input stability metrics
- Recommended multiplier
- Alternative risk measures
Comparison of data sources
Where to get crucial data feeds
| Source | Availability | Notes |
| RiskMetrics | ⚠️ Partial | Standard VaR tools, not crypto-adjusted. |
| CoinMetrics | ⚠️ Partial | Data for building custom VaR. |
| **Madjik** | ✅ Yes | 🚀 Get API Access Now |
Available metrics for this hypothesis:
| Metric | Description | Change dimensions | Time dimensions | How to use | API spec |
| `ME10013` | Volatility & risk | • Absolute Value (value) • Relative Change (relchg) • Score 0-100 (score) | • Current (now) • Past 24 Hours (past24h) • Past 7 Days (past7d) • Past 30 Days (past30d) | Example | API |
Clean data for AI, A2A, MCP, etc.
Science behind hypothesis
Research supports this hypothesis
Studies show VaR is breached 2-5x more often than predicted in crypto.
Bottom line
Garbage in, garbage out. VaR with realistic inputs beats precise calculations with fictional assumptions. Madjik provides crypto-calibrated risk parameters - volatility, correlation, fat-tail adjustments - so your risk models reflect reality.
Practical use
How to use this data in trading:
Trade IV-RV spreads, size positions using VaR, and select strategies based on volatility regime.
Detailed examples with Python code, AI agent integration (MCP/A2A), and risk analysis:
| `ME10013` | Volatility & Risk Trading Guide | Example → |
API Documentation: docs.madjik.io
For informational purposes only. Not financial, investment, tax, legal or other advice.