Options Structure Rules
The options structure family is settled by episode evidence — only the signal (ranking / weighting / filtering) is normally in play. These are hard constraints; a violation is an automatic certification FAIL.
The spread-shape rule (hard constraint)
- Long-dated exposure (≥ ~120 DTE) = outright long calls ONLY. Uncapped convexity, no short leg. No long-dated debit or vertical spreads (e.g. a Mar-'27 ATM/+3/+10/+20 capped vertical is a violation and cannot certify).
- Any debit / vertical spread must be short-dated — expiring in ≤ ~30 DTE. Thin is fine there.
Re-confirm compliance on every finalist by reading leg DTE/strikes from get_portfolio. Violation =
automatic FAIL, no exceptions.
Short-dated verticals are candidates, not banned
A prior thin +3%-wide, uncapped, every-name-eligible ≤30-DTE sleeve FAILED cert (theta bleed, fee churn, cratered the April-2025 fold). That disproved one point, not the class. A short-dated wide, share-capped debit spread is a different structure — let the sweep + OOS cert adjudicate it, but instrument it so the failure mode can't hide: width is a gene, sleeve share is a capped gene, and it must clear the same OOS gates plus a turnover/fee guardrail (report per-fold fees + turnover + sleeve share; a prior sleeve ballooned fees to $1.5–8.9k/fold vs the incumbent's ~$100–600 and must survive the April-2025 fold).
The settled structure family
Outright-calls-only momentum-LEAP: a 7-rung deep-OTM ladder (ATM / +10 / +20 / +35 / +50 / +75 / +100% OTM, 365–730 DTE), TP +250%, affordability backfill, SelectTop 21. Alt-data may touch the rank signal, entry filters, per-name tilts, or an overlay condition — never the structure rungs, the spread-shape rule, or the universe.
The take-profit convexity-cap footgun
If multiple "always" take-profit closes exist (e.g. P/L ≥ 20% / 50% / 200%), the lowest one binds first and effectively caps every winner at ~+20% — defeating the uncapped-convexity rationale for outright calls. It is the single most likely reason returns trail a let-winners-run book. Sanity-check the exit ladder for this on any live book; the deploy config uses a single high TP (e.g. +250%) or none.
The affordability ladder (structural feasibility)
At $25k cold start, a single-template ~$1k-per-name book structurally fails. Minimum defined-risk structures on the expensive tail (META / LLY / GS / TSM / ADI / AVGO) cost $1.3–2.4k per contract.
- Every template set must end in a rung the most expensive name can fill — never remove the cheapest rung when sweeping structures.
- Order templates expensive→cheap so the engine fills the first affordable rung.
- Cheap structures as the primary mechanism = kill path; the ladder must give every name an affordable unit.
totalBudget ≥ K × perNameAllocationper tier. Budgets cap cost basis, not market value — always verify on the tape withaudit_backtest_posture. Dollar budgets do NOT de-risk like %-of-NAV sizing.
See breadth-audit for measuring whether the ladder actually participates at fixed $25k.
Known losers on this book (don't re-test without new cause)
- Stop-losses on long calls / hold-to-expiry LEAP books — whipsaw; tested twice, hurt twice.
- SPY > 200-SMA regime gates — go flat in 2022 → can't clear the certification activity floor; a 200D-SMA momentum filter also tested WORST.
- Long-dated verticals — banned by the spread-shape rule.
- Thin short-dated sleeves at uncapped share — theta-bleed FAIL (see above).
- Narrow ranked rotation (top-K ≤ 5 with shared cadence) — keep K > 5.
Cadence note: DaysSinceLastRebalanceOptionOrder / DaysSinceOrder — null >= N is TRUE on fresh
books; the cadence clock is shared across RebalanceOption strategies.