Two-Sleeve Book — Income Engine + Trend, validated on real marks

The session's headline, in one place: sell defined-risk premium (Iron Condor) as a standalone income sleeve, and run the trend signal as a separate sleeve. Because the two are weakly correlated, the 50/50 book has a lower drawdown than either component alone — exactly the "small DD" goal. Income engine confirmed on real BTC+ETH Tardis marks.

Core idea. Don't overlay options on the same future trade (that's just leverage or a covered call — r/DD unchanged). Instead run two independent sleeves: (A) the trend future (signal-direction), (B) a weekly defined-risk Iron Condor harvesting the volatility-risk-premium. Their bad weeks don't coincide → combined drawdown drops below both.

1 · Income engine — Weekly Iron Condor on REAL Tardis marks (2026)

Sell 2.5% / buy 8% on both sides, hold to expiry. Lim = patient mid fills, Exec = cross the spread (worst case). Positive even at full cross; defined wings cap each week (~−3% max) → small DD. Real bid/ask spread ~7-8% but the credit covers it.

Asset · Fill	Σ P&L%	per week	Win%	MaxDD	Ø credit	Ø spread	weeks
ETH · Lim (@mid)	+19.5	+1.30%	80%	−2.7	2.58%	7%	15
ETH · Exec (cross)	+12.2	+0.82%	73%	−4.2	—	—	15
BTC · Lim (@mid)	+8.3	+0.49%	71%	−5.0	2.08%	8%	17
BTC · Exec (cross)	+5.1	+0.30%	71%	−5.2	—	—	17

Real Deribit marks, ETH & BTC, 2026. ETH pays ~2.5× more than BTC — higher IV = fatter premium (not spread tightness). Source run_nik_ic_tardis.py.

2 · Two-Sleeve on REAL marks (ETH 2026, 15 weeks)

Sleeve	Σ P&L%	MaxDD%	r/DD
A · EXPANSION future (move%)	+26.1	−5.7	4.6
B · Weekly Iron Condor (real)	+19.5	−2.7	7.1
A + B · 50/50	+22.8	−2.1	11.1

Combined DD (−2.1) is below both sleeves (−5.7 / −2.7); r/DD (11.1) above both (4.6 / 7.1) — on real marks. corr(A,B) = +0.45 (positive, yet the worst weeks don't coincide → averaging still cuts the DD). Source run_nik_two_sleeve_real.py.

3 · Two-Sleeve across all algos (Model 2021–2026, weekly)

Trend sleeve A	corr(A,B)	A alone (Σ/DD/rDD)	A+B 50/50 (Σ/DD/rDD)
MA_CND_CROSS H4	−0.30	593 / −16 / 37	358 / −8 / 47.5
EXPANSION H4	−0.06	1122 / −11 / 101	622 / −6 / 97
STOCH H4	+0.10	957 / −56 / 17	540 / −28 / 19
MA_CND_CROSS H1	−0.02	413 / −57 / 7	268 / −27 / 10
RISE_FALL H1	−0.02	339 / −51 / 7	231 / −24 / 10

Sleeve B = weekly Iron Condor (defined risk). Combined halves the DD for every algo; r/DD equal-or-better. The managed (defined-risk) IC is key — a naive always-on strangle (DD −58) dilutes instead. Source run_nik_algo_credit.py (two-sleeve).

Honest caveats. Real-marks windows are 2026 only (~15–17 weeks, one bear) → the per-week rate (+1.3%/wk ETH) is regime-flattered (high IV = fat premium; the model's full-history average is ~+0.46%/wk). The two-sleeve corr (+0.45 real / ~0 model) is noisy on 15 weeks. What's structural & robust: defined-risk IC harvests real premium with a capped tail (survives full cross-fills on BTC & ETH), and combining it with a decorrelated trend sleeve lowers the book drawdown below either component. What's regime-dependent: the absolute rate.

Sources: run_nik_ic_tardis.py · run_nik_two_sleeve_real.py · run_nik_algo_credit.py · build_two_sleeve_page.py. Detail on the credit-spread family: 💰 Credit Spreads · signals: 🤖 Nik Futures-Algos · all strategies: 📊 Scorecard.