Rolling-income covered call on a held ETF (buy-write pillar)

ID: S033 Slug: roll-income-covered-call-etf-v1 Run date: 2026-07-02 Stage: Stage 1 (modeled screen — forward-pending) Outcome: SPLIT, and a low-confidence one. DIA Δ30/Δ20 mechanically clear the frozen gate; QQQ Δ30/Δ20 are cut on the absolute 50% drawdown ceiling. But the DIA "pass" is dominated by a flat-IV modelling artifact — the modelled overwrite appears to beat the held index by ~5pp/yr, which is not credible for a covered call, and a realistic skew haircut removes most of it. So no edge is claimed: this is forward-only, not live, not promoted. Headline metric: flat-IV excess +5pp CAGR → +1.5pp after a 3-vol-pt skew haircut (skew-sensitive)

Positioning — honestly. Same low-alpha, risk-reducing (not tail-protected) character as S032, applied to a permanently-held ETF. On this modelled screen no edge is claimed — the apparent gain is largely a flat-IV artifact (see the skew-haircut exhibit). It is not low-risk in absolute terms: the full downside is retained, and on QQQ the permanent-hold tail (−82% in the dot-com crash) is exactly why those cells are cut. Forward-only; nothing live or promoted.

What we tested

The continuous / rolling variant deferred by S032. Where S032 writes one call and closes the whole position at expiry (episodic), S033 keeps the ETF permanently and writes a new ~30-DTE call every cycle — the covered call marketed to retail as "income". The core is established on the S027 RSI-2 pullback with a 4-unit cap and never sold (so the first few dips build the holding, after which entry timing is near-vacuous — a permanent holding is nearly start-date-independent by construction). Each cycle: write a call at the cell's delta, hold to expiry, keep the unit, cash-settle if in-the-money, immediately re-write. Fully invested / reinvested (BXM-style total return), net of friction on every write. The option leg is modelled (Black-Scholes fed with VXD/VXN as ATM IV). Matrix: DIA/QQQ × delta {30, 20} — four cells, four pillar trials (cumulative → 10).

Gate (frozen, risk-adjusted): Sortino(overwrite book) > Sortino(held-ETF baseline), max-drawdown (DD) ≤ the held baseline and ≤ an absolute 50% catastrophe ceiling.

What we found

Two things are true at once. Mechanically, the DIA cells clear the frozen gate — higher Sortino, lower drawdown than just holding. But the magnitude is not trustworthy: the flat-IV overwrite beating the index by ~5pp/yr contradicts every real covered-call index (BXM/BXY historically match or lag the S&P), and the skew-haircut column shows why — most of the excess is premium the model over-collects because it prices OTM calls at ATM volatility, every cycle. The residual after a modest haircut (+1.5pp, Sortino 0.53→0.64 on DIA) is within the plausible variance-risk-premium range but swings wildly with the assumed skew, which only real chains can settle.

What we learned

The modelled result is dominated by a flat-IV artifact: pricing every out-of-the-money call at the at-the-money volatility index over-collects premium, and over hundreds of rolling cycles that error compounds into an implausible ~5pp/yr edge over the held index — which contradicts every real covered-call index (BXM/BXY historically match or lag the S&P). A modest 3-vol-point skew haircut removes about two-thirds of it (DIA Δ30 falls from +5.3pp to +1.5pp). The residual is within the plausible variance-risk-premium range but swings wildly with the assumed skew. The clean lesson: a rolling covered call cannot be validated on a flat-IV model — the modelling error is larger than the effect it is meant to measure.

Why QQQ is cut

The gate's absolute 50% drawdown ceiling exists precisely for this. A permanently-held QQQ position spanning 2000–02 drew down 82%; the monthly call trims that to ~68–73%, still far past a level any allocator would hold through. The call cushions by roughly the premium — it is not tail protection — so it cannot convert an uninvestable buy-and-hold tail into an investable one. That is the honest structural limit of the covered call, and it is why the left-tail overlay (collar / protective put) is a separate, distinct structure on the roadmap.

What this doesn't tell us yet

Because the surviving edge is skew-sensitive — it swings from "beats the index by 5pp" to "roughly matches with a small risk-adjusted tilt" under a plausible haircut — it cannot be pinned down without real option chains (real implied volatility, skew, fills, rolls). So this modelled screen says nothing reliable about the after-skew magnitude; only a forward test on real chains can settle it.

What happens next

No edge claimed. The rolling-income covered call cannot be validated on a modelled screen — flat-IV overstatement compounds every cycle and dominates the result, and a plausible skew haircut is enough to swing it from "beats the index by 5pp" to "roughly matches with a small risk-adjusted tilt". The DIA cells' mechanical pass moves to forward-paper (Stage 3) with real chains — real IV, skew, fills, and rolls — where the premium is what the market actually pays, not what flat ATM IV imagines. The QQQ cells are cut. Nothing here is live or promoted.