Every trading day the engine computes a rolling historical volatility (HV) proxy for each ticker in the universe using 21 trading days of realized returns. That HV is scaled by 1.15 to approximate implied volatility, then compared against all tickers in the same GICS sector.

The Z-score measures how many standard deviations the ticker's current IV sits above or below its sector peers — not above or below its own history. This cross-sectional design means a single name in a low-vol sector triggers the gate even when its absolute IV looks normal.

Sector groups with fewer than 3 members roll up to the market-wide pool. Z-scores are computed fresh each day and form the input to the three-tier gate below.

Income-Focused uses Precursor Intelligence's first-generation binary gate: full deployment when IV is normal, complete pause when compressed (Z < −0.5σ). Capital-Protective upgrades this to three tiers — preserving income generation during borderline compression while only fully pausing at genuine extremes (Z < −1.0σ).

In REDUCED mode the put is written shorter-dated with a lower delta (less premium, less assignment exposure). This replaced the 2017-style scenario where the binary gate caused a near-zero income year. With the three-tier gate, borderline compression days still contribute income — just at reduced scale.

The PAUSED threshold at Z < −1.0σ represents genuine IV compression — the market is structurally under-pricing risk and no new exposure is appropriate.

The IV gate addresses vol compression but not sustained directional decline. In a bear market, IV is often elevated (the gate stays ACTIVE or REDUCED) even as the underlying falls steadily. An ACTIVE IV signal does not mean the stock is safe to put.

The trend filter adds a second veto: no new CSP is written on a ticker that is currently trading below its 200-day simple moving average. The 200-day MA requires at least 50 prior trading days to warm up; before that, the filter abstains (no block, no write).

The filter fires after the IV gate. A PAUSED ticker never reaches the trend check. A REDUCED-tier ticker still runs through the filter — if it is below its 200-day MA, the write is also blocked.

This addresses the 2007–2008 failure mode where elevated IV (due to crisis volatility) would have kept the gate ACTIVE while the underlying collapsed 50–70%. The trend filter would have blocked writing on any name in a confirmed downtrend, regardless of what IV was doing.

The first three layers filter individual ticker risk. The Granger causality network adds a systemic view: it detects when one ticker's past returns carry statistically significant predictive information about another ticker's future returns.

The engine runs pairwise Granger causality tests across the ticker universe using a lag window of 1–5 trading days. A pair passes if the F-test achieves p < 0.05 after a Bonferroni correction for multiple comparisons. Significant pairs are mapped into a directed graph where an edge A → B means "A's returns reliably precede B's returns."

Tickers with many outgoing edges (high out-degree) are leaders — movements in these names tend to propagate through the rest of the universe hours to days later. Leaders get elevated monitoring weight. If a leader is in a downtrend, it is a stronger sell-off signal for follower names than the follower's own price action alone.

The network is rebuilt on a rolling 90-day window so relationships that have disappeared (de-correlated markets) naturally drop out.