The BTC Calendar Strategy (CAL_BTC_MA) — in detail

A direction-neutral double calendar, triggered by a trend algo. The only genuine diversifier in the book (~0 correlation to everything else). What it is, why it earns, the numbers, the full trade-by-trade flow, and the key question: can we trade bigger size?

1. The structure

Trigger:   MA_CND_CROSS_H4 trend signal fires on BTC (long OR short — direction irrelevant)

Open:      BUY  28-day call ATM  +  SELL 14-day call ATM   (same strike)
           BUY  28-day put  ATM  +  SELL 14-day put  ATM   (same strike)
           = 4 legs, all at the current price (ATM)

Hold:      ~13.5 days (just before the 14-day options expire)
Close:     all 4 legs

You pay a net DEBIT of ~2.88% on entry (= your maximum loss).

2. Why it earns (two effects)

Theta differential: the sold 14-day option decays faster per day than the bought 28-day. After 13.5 days the short is nearly worthless while the long keeps ~14 days of value → the difference is profit, as long as spot stays near the strike. Pinning: ~2 weeks after an MA_CND_CROSS signal, BTC tends to sit back near the entry level (|move| 5.9% vs 7.3% random) — exactly what the calendar needs.

 Profit ▲        _________
        │      /           \          ← max profit when spot ≈ strike (pinning helps here)
       0│____ /             \ ____
        │  trend away = loss (max = debit 2.88%)
        └──────────┬──────────────► spot after 14 days
               entry price (strike)

Direction-neutral: because it is call AND put (double calendar), it does not matter whether the signal was long or short — measured: SAME +0.493 vs OPP +0.491 = identical. The signal is only the timing.

3. The numbers

Model positive every year (2024 +74 / 2025 +65 / 2026 +30, % of notional summed). The 12 real trades CONFIRM the edge survives real spreads but are too few to size on. Realistically expect the model values (~+1%/trade, 67% win); the +1.55%/91% is small-sample luck.

4. Every trade in detail (the flow) — 181 trades, hold-to-expiry

Each trade: entry, signal direction, spot, the two ATM strikes (put/call), net debit (= max loss), the spot path during the 13.5 days (+max up / −max down — shows whether spot ran away and came back = the pinning), the net move to exit, the PnL and the cumulative curve. Note: this calendar does NOT reposition — it is opened and held (flow: OPEN → EXPIRY). A repositioning variant would show extra intermediate events here (see the Shifting-IC page — tested there, does not work).

Three trades fully broken out (all 4 legs)

5. Execution & management — when to trade, and does it need adjusting?

Time-sensitivity: NO — barely any edge lost up to ~1–2 days delay

Tested: entry delayed by X hours after the signal (at the then-current spot), otherwise identical:

Up to ~24–48h delay keeps essentially the full edge; positive every year. Only beyond 3 days does it fade (still positive). Reason: the edge is a slow effect (2-week pinning + theta over 13.5 days) — seconds/hours don't matter. → comfortably tradeable by hand, patient limit fills possible.

Management: simple beats everything — no TP, no stop, no hedge

Management variants vs the simple hold-to-13.5d (M2 24-26, MA_CND_CROSS, 179 trades):

Conclusion: the strategy needs no active management. Take-profit cuts the winners, stop-loss cuts the pinning recoveries, delta-hedging 5×'s the worst week (the calendar is short gamma near the 14d expiry — hedging buys high/sells low), and "rolling" is just more volume with a weaker 2nd cycle. Open → leave 13.5 days → close. Combined with the time-insensitivity: fully passive, hand-tradeable, no adjusting needed. (Note: naive re-centering / shifting the strikes onto the new spot HURTS, because the edge is reversion to the ORIGINAL entry level — shifting abandons exactly the level price returns to.)

6. Can we trade BIGGER size? — the honest analysis

Practical sizing: being defined-risk, size via a max-loss budget. E.g. max 3.62% loss per calendar, up to 9 concurrent → absolute worst case (all at max loss simultaneously, very unlikely as they are staggered) ~33% of per-trade notional. The sleeve drew down ~−9% standalone (1-BTC notional over 2.5y). Pick per-trade notional so even the cluster case stays inside the book risk budget (e.g. sleeve max loss ≤ 5-8% of total book).

7. Risk & honest caveats

8. Next step — find the size via live paper

Sources: run_grand_portfolio.py, run_nik_cal_real26.py, run_nik_cal_boot.py, run_cal_sizing.py, run_cal_ledger.py. Model M2, indicative. German version: kalender_strategie.html.