How to manage SOLD weekly options (Fri→Fri, BTC/ETH). Tested on real Tardis marks (18 weeks 2026) AND a 3-year model history (124 weeks, by-year) — every rule below survived both. Companion to the overnight report and the 5-sleeve portfolio.
| Sold weekly PUT, exit day sweep | Total | MaxDD | r/DD | note |
|---|---|---|---|---|
| BTC real 2026: HOLD to expiry | +5.7% | −19.3% | 0.3 | tail hits late-week |
| BTC real: CLOSE WEDNESDAY | +14.7% | −5.4% | 2.7 | return ×2.6, DD ÷3.6 |
| BTC model 24–26: HOLD | +74.0 | −40.2 | 1.8 | by-year +58/+15/+1 |
| BTC model: WEDNESDAY | +67.1 | −11.7 | 5.7 | +31/+21/+15 — every year, smoother |
| ETH model: HOLD → WEDNESDAY | +33.0 → +48.6 | −55.1 → −35.5 | 0.6 → 1.4 | return UP and DD down |
Why it works: by Wednesday the sold put has earned ~77–80% of its theta (median), while the remaining two days carry most of the tail (late-week / weekend crashes). Mon-exit is too early (theta not yet earned, negative), Tue is okay, Wednesday is the sweet spot, Thursday already degrades. Operationally trivial: buy back Wed 08:00 UTC.
| BTC PUT + Wed exit (model 24–26) | Total | MaxDD | r/DD | verdict |
|---|---|---|---|---|
| naked + Wed | +67.1 | −11.7 | 5.7 | the baseline is already good |
| + long put −5% (spread) | +11.3 | −7.4 | 1.5 | hedge eats 80% of the edge — only if margin requires defined risk |
| + futures-stop @ breakeven, HOURLY | +72.6 | −10.2 | 7.1 | best across regimes (ETH: r/DD 1.4→4.1, DD −56%) |
| + stop checked once DAILY | +50.6 | −17.6 | 2.9 | worse than NO stop — late reaction, deeper entry, same whipsaw |
Honest conflict: in the real 2026 window the hourly stop whipsawed (V-shaped dips before Wednesday; the real crashes came late-week — which the Wed exit itself already avoids). Across 3 model years the stop pays. If you run a stop: continuous/hourly monitoring with auto-orders — a manual daily check is strictly worse than nothing.
| PUT | CALL | |
|---|---|---|
| Exit | close Wednesday (theta early, tail late) | hold to expiry — Wed exit destroys it (ETH real +31.4 hold vs −1.2 Wed; theta earned LATE) |
| Protection | hourly BE-stop optional | BTC: hourly BE-stop mandatory (naked −3.0/−48.4 over 3y; stopped +41.6/−18.8). Spread eats everything. |
| ETH | ✓ trade it | ✗ don't sell ETH calls at all — nothing makes them positive over 3 years (naked −4.6/−67.7, stopped −11.4, spread ~0) |
| Variant | BTC M2 24-26 (by-year) | ETH M2 24-26 (by-year) | real 2026 exec (BTC / ETH) |
|---|---|---|---|
| ATM/+5% | −12.8 / DD −27.6 (−24/+10/+1) | +0.3 / −17.6 | +1.0 / +2.9 |
| 2%/+5% | −3.7 / −23.3 | +8.0 / −11.8 | +3.1 / +3.7 |
| 2%/+7% | −4.1 / −28.9 | +13.4 / −17.7 r0.8 (−11/+16/+8) | +5.3 / +9.1 |
| 3%/+8% | −8.4 / −29.0 | +13.9 / −17.3 r0.8 (−13/+17/+10) | +6.7 / +8.1 |
Verdict: (a) BTC: spreads LOSE across 3 years (2024 bull kills every variant, −20 to −24) — the validated BTC call route stays naked + hourly BE-stop (+41.6/−18.8/r2.2, positive every year), clearly superior to any spread. (b) ETH: wide spreads (2%/+7%, 3%/+8%) are the FIRST ETH-call structure with a positive 3-year total (+13.4/+13.9) — naked (−4.6) and stopped (−11.4) both lose. BUT r/DD is only 0.8 and 2024 is still negative (−11/−13): it upgrades ETH calls from "never" to "weakest corner, only if you insist" — the book rule (don't sell ETH calls) stands. Real-2026 numbers (+8–9% exec) are bear-flattered.
3c. Optimization matrix 2025+2026 (5 strikes × 3 widths, 2024 shown as robustness reference): the consistent pattern across the whole matrix — not one lucky cell — is further OTM short (2–3%) + TIGHT width (+3%): the tight cover hard-caps bull damage while the OTM short still collects. Best cells: ETH 3%/+3% → 25+26: +19.2 / DD −6.7 / r2.88 / win 79 / worst −2.5, and 2024 only −7 — the first call-spread that nearly survives the bull year (3y total ≈ +12). BTC 2%/+3% → 25+26: +16.8 / DD −6.7 / r2.49 — but 2024 −20: for BTC the naked+hourly-stop route stays superior across regimes; the tight spread is the defined-risk alternative when margin matters. ⚠️ These cells were SELECTED on 25+26 (15 candidates) — expect out-of-sample below these numbers; the structural tilt (tight + further OTM) is the durable part, not the exact cell.
| Mode (call-spread base from 3c) | 25+26 Total | DD | r/DD | worst wk | 2024 | real 2026 exec |
|---|---|---|---|---|---|---|
| BTC call-spread SOLO | +16.8 | −6.7 | 2.5 | −2.8 | −20 | +3.1 / r0.7 |
| BTC + put-spread after +2% rally ★ | +27.4 | −4.3 | 6.4 | −2.2 | −11 | +9.9 / r3.9 |
| ETH call-spread SOLO | +19.2 | −6.7 | 2.9 | −2.5 | −7 | +4.6 / r1.4 |
| ETH + put-spread after +2% rally ★ | +23.8 | −4.3 | 5.5 | −2.2 | −7 | +7.6 / r5.7 |
| trigger at short-strike TOUCH instead | worse everywhere (ETH even below solo) — too late/too rare | |||||
| STATIC iron condor (both sides upfront) | good in the 2026 bear, weaker across regimes (win 46–50%) | |||||
Rule: open the bear-call-spread Friday; if spot rallies +2% intra-week, sell a put-credit-spread (2% OTM from the NEW spot / 3 points wide, same expiry), hold both to expiry. Improves EVERY column vs solo — return, DD, worst week, even 2024 (BTC −20 → −11) — so it is not period-picking. The feared double-loss (rally → add puts → reversal) does NOT materialize: after a +2% rally the new put strike sits ~at the old entry level, so a reversal first rescues the call side before it threatens the puts; defined risk caps both ends. Mechanically the same medicine as the sequential diagonal (momentum gate on the put side) — second independent confirmation of that pattern. Operationally: one price alert at +2%, no monitoring. Caveats: +2% is one tested trigger (touch variant is worse → some parameter sensitivity); real n=18.
For the 2wk/1wk double diagonal: selling the CALL side first and adding the PUT side only after spot rises +2% (a momentum gate on the dangerous side) cuts model MaxDD −22% (−21.6 → −16.8) at r/DD 4.5 → 5.0, costing −13% return (puts deployed in 76% of campaigns). The first sequencing trigger that ever helped. CALL_ONLY is NOT the answer (+42.5/−21.8 — the DD does not come from the put side). Model-only so far; real-marks check pending.
run_sven_weekly_exit.py · run_sven_wed_hedge.py · run_sven_seq_diag.py ·
run_sven_theta_decay.py · log: NIK_ALGOS_LOG.md.