Overnight Report — BTC/ETH Weekly Premium Book (the honest conclusion)

TL;DR for the morning

After mapping the whole surface (strikes, tenor, management, execution) on a regime-robust Monte-Carlo (800 bootstrapped paths) + real Tardis order-book data, the honest conclusion: no crypto weekly-premium structure is both high-return AND low-drawdown after real costs. You pick ONE:

Low drawdown → the hedged far-OTM diagonal (3wk long / 5-OTM short / close-on-ITM): ~+13%/yr (~1%/mo), median DD −4.7%, p95 −12%, win 61%. The tail is guaranteed-capped by the long wing. With good (limit-near-mid) execution: ret/DD ~3.4–4.3.
More juice, accept the tail → naked weekly selling: ~+29%/yr, win 68–77%, but median DD −14 to −19%, worst −60 to −71%. And a stop does NOT save you (see below).

Recommendation: if your stated priority (small drawdown) holds — trade the hedged variant, small, with limit orders near mid, sized to your DD budget, as a low-correlation diversifier to the trend book. Start with ONE small position to measure your real fills before committing size.

1. The fork — hedged vs naked vs naked+stop (Monte-Carlo, 800×1yr, 5% spread)

variant	ret med	p5	DD med	worst	P(DD<−10%)	win
Hedged 3wk / 5-OTM / close ★	+13	−3	−4.7	−22	9%	61%
Naked weekly 3-OTM	+30	−11	−16.0	−64	84%	72%
Naked weekly 4-OTM	+29	−7	−13.9	−61	73%	77%

2. The key mechanic — a HEDGE caps the tail; a STOP fails on gaps

A naked-selling stop-loss looked like the holy grail (+48%/yr at −7.7% DD) — but that assumed a clean fill at the stop. Modelled realistically (intra-week trigger + slippage, because crypto gaps), it collapses:

stop model (naked 3-OTM)	ret med	DD med	worst	P(DD<−10%)
stop @1.5× clean (optimistic)	+48	−7.7	−30	29%
stop @1.5× realistic (gap+5% slip)	+20	−16.5	−64	86%
stop @3× realistic	+26	−16.8	−67	87%

Why: crypto gaps through stops (you fill far past the trigger), and whipsaws stop you out then reverse. A long option wing, by contrast, pays its intrinsic no matter how big or fast the move — it's a guaranteed cap, not a hopeful one. This is the whole reason the hedged variant is the real low-DD choice.

3. The chosen variant — exact rules (BTC & ETH)

Every Friday, only if DVOL < its 90-day median (IV filter).
Buy 3-week long: call +5 strikes OTM, put −5 strikes OTM (cheap far wings).
Sell 1-week short: call +4 OTM, put −4 OTM (4-OTM strangle, 1 strike inside the long).
Day 7: roll the weekly shorts to the new 4-OTM. Hold the long.
Close rule: if a long goes ITM (a big move reaches +5/−5), close BOTH longs — lock the gain, go flat (rare → cheap).
Else day ~21 cash-settle, re-enter. Do NOT roll the winner / use a trend filter / run naked.

BTC grid $1000, ETH grid $25. (2wk long is essentially equivalent — within MC noise.)

asset	M2 25+26	M2 DD	Tardis Exec 2026	Tardis DD
BTC	+23	−4.3	+4.7 (full-touch)	−1.3
ETH	+28	−8.4	+6.5 (full-touch)	−1.7

4. Execution is the real lever (not more tuning)

fill quality (BTC chosen variant)	ret med	DD med	ret/DD
Mid (patient limit @ mark)	+17	−3.9	4.3
Half-touch (~2.5% RT — realistic)	+15	−4.3	3.4
Full-touch (~5% RT)	+13	−4.7	2.7

The gross edge is the variance-risk-premium (~5–10%/yr in crypto); the bid-ask eats ~half. So limit orders near mid roughly add a full point of ret/DD. Discipline at entry > any further parameter tuning.

5. Sizing (DD scales linearly with size)

At 1× the reference, the hedged variant's 95%-worst-year DD ≈ −12% (median −4.7%). Linear: for a −6% worst-year budget run ~0.5×; for −12% run 1×. Median return scales with it (~+13%/yr at 1×). Because the structural DD is low, you can size up toward 1.5–2%/mo and still keep the worst-year DD in the teens — something the naked book can't do (its tail is −60%+ at any size).

6. Honest caveats

Negative skew + bootstrap limit: the MC can't draw a tail bigger than 2024–26; a true black-swan hits the thin far-OTM premium harder than the modelled −22%. Size with buffer.
Tardis 2026 is one (benign) regime & ~4 months; the M2 2024+ and MC are the conservative truth, not the rosy −1% real DD.
It's a diversifier, not a star: ret/DD ~3.4 standalone is modest but uncorrelated to your trend book → portfolio value.
Execution is the make-or-break: cross the full spread every leg and ret/DD drops to ~2.7; this only works with limit discipline.

Built overnight. Sources: run_montecarlo / run_mc_strikes / run_mc_grid / run_mc_spread (DD distributions), run_naked_weekly + run_naked_stop_real (the fork & the stop reality-check), run_final_variant (BTC/ETH hedged), run_diag_final (tenor/strike/rule sweep). Deep-research literature section appended when the workflow completes.