Take the best trend signals (Donchian / HHV breakouts, MACD-aligned โ the filter from
the combo study) and, instead of trading the future, buy a cheap directional option in the
signal's direction. We only keep the entry windows that survived the time study โ DAY + WEEKEND
(night was a confirmed loser on real prices). For each signal we test five structures over three
expiries, price them two independent ways, and split LONG vs SHORT and by year.
How to read
- R = return on the net debit you pay. +1.0 = the option doubled (you got 2ร the premium back);
โ1.0 = total loss of the premium. It is not % of account โ premiums are small, so this is a
lottery-ticket-style sleeve, not a core book.
- M2 = Model-2 (our vol surface, bias-corrected) โ the broad 2024+ history.
Tardis = real Deribit marks โ the 2026 reality check (~4 months, small n).
- Year columns (24/25/26) = M2 cumulative R per calendar year at the ~1-week expiry.
- LONG signal โ buy a CALL; SHORT signal โ buy a PUT. Reported separately.
The structures
- Long ATM / 3% OTM โ outright option. Full directional upside, you pay the whole premium.
- Vert spread 5% / 8% โ buy ATM, sell an OTM option that far away. Cheaper, but upside is
capped at the short strike.
- Butterfly 5/10% โ buy ATM, sell 2ร at +5%, buy +10%. Cheapest, but profit is pinned
near the +5% target and collapses if the move overshoots โ wrong tool for a trend that runs.
- Expiry ~1 / 2 / 4 weeks โ shorter = cheaper premium and higher return-on-premium; the trend
move is fast, so longer expiries mostly just pay extra theta.
Headline findings:
- The SHORT side (puts) carries the book โ and it's the robust one. Donchian SHORTโput, day,
short-dated outright: M2 +1.0โ1.4 (ret/DD 13โ15), real Tardis +1.7โ2.2, positive every year.
The LONGโcall side is weaker and fragile in 2026 (HHV calls turn negative).
- Shortest expiry (~1 week) wins (cost ~2% of spot vs ~6% at 4 weeks).
- Outright (ATM / 3% OTM) > vertical spreads > butterfly. The butterfly caps the trend's
fat tail and goes negative at longer expiries โ don't use it for trend bets.
- Night excluded โ cheap premium โ edge; it lost on real marks.
Caveats: Tardis is only ~4 months of 2026 (small n, flagged *) โ read
it as direction, not precision. M2 is the long history but blind to intraday premium. Return is on the
premium (small absolute size).