| name | vol-regime |
| description | Code patterns for classifying the current volatility regime and detecting anomalous recent moves. Always load alongside fetch-yfinance — the examples require a yfinance data fetch at the top of the same script. Load references/examples.md for working code. |
Volatility regime classification
What this skill provides
references/examples.md — Working code patterns for:
- Pattern 1: Rolling 30-day annualised vol + regime classification (low / normal /
elevated / extreme)
- Pattern 2: Anomaly detection — z-score of the most recent daily move
- Pattern 3: Adaptive trend window selection based on regime and anomaly signals
These patterns are designed to be combined with a data-fetch block in a single
code execution. Do not call run_code separately for data fetching and regime
classification — combine them.
Typical usage
Load fetch-yfinance and vol-regime, read both references/examples.md files, then write
one complete block that fetches the data and computes the regime.
Regime thresholds (WTI crude oil)
| Regime | Annualised vol (%) |
|---|
| low | < 20 |
| normal | 20 – 35 |
| elevated | 35 – 55 |
| extreme | > 55 |
Adjust thresholds for other assets. These are calibrated to WTI's historical
vol distribution (2020–2025 median ≈ 31%).
Output of Pattern 3
Pattern 3 returns a trend_window integer (15 or 30 days) that you should
pass directly to the trend-projection skill's fitting step.
