S-048 — return-stacking-tbill-neutral-0dte-v1 (T-bill base + market-neutral overlay + defined-risk 0DTE)

ID: S-048 Slug: return-stacking-tbill-neutral-0dte-v1 Type: OVERLAY (portfolio-construction study — not a market edge) Grid: internal (members-only exploration page, not on the public /lab grid until frozen) Status: 0DTE SLEEVE FAILED — 2026-07-20 settle-bug audit. This construction study stacked the S049 dual 0DTE fly onto a T-bill base + S038 neutral overlay — but the 0DTE fly sleeve was a settlement-proxy artifact (settled off SPY×10; corrected → no edge), so its contribution to the stack is void and the stacked-return illustration below is superseded. The T-bill + S038 sleeves stand on their own. Was: REGISTERED / PRE-REGISTRATION 2026-07-13, a construction study (not a signal), NOT frozen, NOT a claim, illustrative-only pending Tasty margin mechanics (research/tasty-portfolio-margin-questions.md). System: Return stacking (proposed — not yet on the public grid). Promotes: candidate lab/candidates/return-stacking-factor-on-beta-v1.md.

The three sleeves (each an existing, independent track)

AMENDMENT 2026-07-15 — 0DTE sleeve upgraded to the S-049 dual-fly. The 0DTE component is now the S-049 book (1 SPX + 5 QQQ) on the complete SPX (+missing-strikes backfill) ∩ QQQ data, not the SPX-only S-045 ×2. Effects: window extended to 2023-01→2026-04 (40 months, vs 21); S-038·0DTE correlation −0.03 (vs +0.21); the 0DTE is now a diversifier at ~25% risk-share (S-038 anchors ~78%), not co-dominant. Sizing is now N_BOOKS (units of the 1-SPX+5-QQQ book), default 1. Results in gates.md. The mentions of "S-045 ×2 / n_contracts / 2022-06→2025-12" below are the superseded original.

Method (frozen before the run)

1. Common monthly window. Study window = 2023-01 → 2025-12 (S-038 holdout ∩ S-045 sessions incl. the new 2024-08→2025-12 OOS). Secondary read: full 2022-06 → 2025-12. Everything at monthly frequency (S-038's native grain); S-045 per-session P&L is summed within each calendar month.
2. Everything expressed as % of the $180k NAV. - Base = DGS1/12 each month on 100% of NAV. - S-038 = its monthly return at the chosen sizing vol (default 10%/yr = the track's native scale; [CALIBRATE] — see below), applied on NAV (a market-neutral overlay needs little net capital). - S-045 = (Σ session P&L in the month × n_contracts) ÷ NAV.
3. Accounting (conservative, never flattering). - Haircut: available buying power = haircut × NAV (default 0.92 [CALIBRATE]). - Financing drag on S-038: subtract (margin_rate − short_rebate) × one_leg_gross ÷ 12 monthly (default spread 1.5%/yr [CALIBRATE] — the single biggest unknown; Tasty Q3). This is the "is the stack free?" term the candidate flags. - S-045 frictions are already inside its P&L (§3a moneyness-aware spread, worse-than-mid fills).
4. Diversification read. Pairwise correlations of the three monthly streams; each sleeve's share of total portfolio variance (risk contribution), not just its mean.
5. Forced-liquidation stress (the tail that matters). A synthetic crisis month where, simultaneously: S-045 posts its capped worst session × n_contracts; S-038 draws −k × its monthly vol (default k=3 [CALIBRATE]); and the PM haircut widens (BP shrinks by stress_haircut, default 25% [CALIBRATE]). Test whether combined BP usage would breach available BP → forced de-lever. Report the worst realised combined month and the stressed month side by side.

Success criterion (pre-registered — a construction test, not an edge gate)

• (1) Stacking helps: combined Sharpe > max(single-sleeve Sharpe) over the window, AND combined max-drawdown ≤ the worst single sleeve's.
• (2) The tail survives the plumbing: the stressed worst-month (§5) stays within the BP buffer at the chosen sizing — i.e. no forced liquidation with ≥ the owner's target buffer free.
• Miss either → the honest read is "the stack is not additive / not safe at this sizing," reported as such (no cut/kill language — this is a construction study).

[CALIBRATE] — owner / Tasty-dependent, NOT filled here

• sizing_vol for S-038 (10%/yr default ↔ the 1.5/1.5 the owner named; confirm the mapping).
• haircut (T-bill marginability — Tasty Q2), margin_rate − short_rebate spread (Tasty Q3), PM % of gross for the equity book (Tasty Q4), target_buffer (owner risk choice), stress k and stress_haircut.
• n_contracts = 2 — RESOLVED by owner 2026-07-13 ("doe maar stack x2 lots"). Owner accepts that at 2 lots S-045 carries ~56% of stacked-portfolio variance (co-dominant with S-038, not a garnish).
• Forward T-bill carry ≈ 4%/yr (owner's current fill); the backtest uses the real historical DGS1.

Skip stages

1, 2 — no historical edge to backtest; this is a construction/recombination validated forward. The sleeves' own edges are validated in their specs (S-038 holdout, S-045 forward). S-048 studies the combination; it goes construction-study → forward paper, never a standalone in-sample gate.

Engine

the exploration workspace · inputs: s038_monthly.json, s049_dualfly_book.json (via the engine), tbill_dgs1.csv · figure fig_stack.png.

How it could fail (must be reported, not hidden)

Correlations that read ~0 can spike toward 1 in a deleveraging crisis; the S-038 financing spread can eat the T-bill carry (killing the "free" stack); and a high-notional 0DTE sleeve can dominate the combined risk, turning a "diversifier" into a concentrated 0DTE bet. The study stresses the tail and reports each sleeve's risk share — it does not sell the mean uplift.

Quant-review caveat (2026-07-15, the exploration workspace): the S-049 dual-fly 0DTE sleeve is an everyday-diversifier, not tail insurance — its diversification is robust to cost/vendor/ commission stress but is fair-weather (SPX/QQQ underlyings ~0.9-correlated → both flies co-lose in a real vol explosion; the dollar tail is not reduced). And the low 0DTE risk-share (~25%) partly reflects recent SPX-fly compression (SR 0.26→0.07 at the 2024-08 data-vendor boundary), not only true decorrelation — so don't read the stack's smooth path as crash protection.