Skip to content

Summary

Amendment target: the VRP axis of the regime definition and the eligibility column of the regime→strategy mapping (v0).

Change: the VRP variable moves from sign to size threshold. Premium-selling structures are eligible on the VRP axis only when

VRP_size = IV30 − forecast_RV30 ≥ θ (in annualized vol points, θ calibrated, initial proposal θ = 3.0)

Between 0 and θ, the VRP axis grants no eligibility by itself: structures may still be proposed, but the burden shifts entirely to the per-candidate EV gate. Below 0, the existing long-vol/defined-risk tilt applies unchanged.

Why (Evidence)

  1. Structural evidence: Chicago Fed WP 2025-17 finds option alphas statistically indistinguishable from zero over ~15 years; Dew-Becker & Giglio document a break ~2012 after which standard S&P option strategies lose significance.12 A sign rule assumes an average premium that no longer exists.
  2. Live journal analogue (process evidence only): the 2026-09-09 rehearsal showed sign-pass / edge-fail — the fit gate passed on a positive VRP, and the attested EV run still failed the edge test. This demonstrates that sign alone is insufficient, i.e. that some size condition is needed. It does not evidence a specific threshold: the run's VRP_size was ≈ 4.85 points (IV proxy 0.1585 − placeholder forecast 0.11), which would pass the proposed θ = 3.0; and the GARCH input was an unfitted placeholder, so the receipt's edge_vs_rv value is excluded from any inference about θ.3
  3. Quorum status: the two studies above plus the rehearsal are rationale supporting a proposal, not a satisfied concordance quorum. Qualification (three concordant journal events or a named historical case-study analogue with explicit reasoning) must be recorded before activation, alongside the calibration and human approval.

Calibration Plan (What Decides θ)

θ = 3.0 is a placeholder. The calibrated value comes from data, computed once the V1 snapshot script (fitted HAR/GARCH forecast — plumbing priority #1) exists:

  1. Forecast series: fit HAR-RV (and a GARCH(1,1) cross-check) on ThetaData/Tiingo data 2016→present; produce daily ex-ante forecast_RV30.
  2. VRP_size series: IV30 (prior close) − forecast_RV30 per day, ex-ante only.
  3. Stratified backtest: premium-selling archetype (45-DTE SPX condor per the attested computation) evaluated per VRP_size bucket (0–1, 1–3, 3–6, >6): net-of-cost EV, hit rate, tail loss, using the attested machinery — never prose arithmetic.
  4. Discipline: purged walk-forward splits, the 2018/2020/2024 stress set mandatory, multiple-testing controls per backtest discipline. θ = the smallest bucket boundary at which net edge turns positive and survives the stress set — or the honest conclusion that no θ works, in which case the amendment dies and the EV gate remains the sole defense.
  5. Documented prediction: the amendment note must predict what changes (fewer eligible days, higher per-trade edge expectation) so the learning function can later score the prediction.

Activation Conditions (All Required)

  • [ ] Fitted HAR/GARCH forecast in production (V1 snapshot script)
  • [ ] Calibration backtest completed and passes the stress set
  • [ ] θ (or "no θ") decided from the calibration output
  • [ ] Human approval recorded in the decision log

Until activation: v0 rules remain in force (VRP sign only), and the EV gate carries the full burden — as it already did on day 1. Caveat: the v0 sign rule is currently computed from proxy inputs (VIX as IV30 proxy; trailing RV, not a fitted forecast), so v0 regime labels are provisional — day-1's vrp+ label used IV-proxy minus trailing RV, not the canonical IV30 − forecast_RV30. Fitted forecasts at V1 recompute the series honestly.

References

Version Ledger

Version VRP rule Status
v0 (current) sign only in force
v1 (this proposal) size threshold ≥ θ, calibrated proposed — pending calibration + approval

  1. Chicago Fed WP 2025-17 ↩

  2. Dew-Becker & Giglio, author-hosted PDF ↩

  3. 96-journal/2026-09-09.md ↩