The 2wk/1wk diagonal (buy 14d far, sell weekly near) is net-LONG vega โ it gains when IV rises and bleeds on a vol crush (which is why call & put both lost in the high-IV-then-falling weeks of early Jan). Two fixes were tested on BTC, M2 2024+, split by entry-IV regime (DVOL vs its trailing-90-day median).
Result:
Caveat: M2 uses one DVOL with no term structure, so the vega magnitude is indicative; the sign and the regime split are the robust takeaways. R = % of notional.
| structure | IV regime | n | Total% | Win% | Worst% | Ret/DD | 2024 | 2025 | 2026 |
|---|---|---|---|---|---|---|---|---|---|
| LONG-vega | all weeks | 109 | +58 | 79% | -7.0 | 4.2 | +31 | +15 | +12 |
| LONG-vega | entry IV LOW (below 90d median) | 76 | +49 | 80% | -7.0 | 7.0 | +24 | +17 | +8 |
| LONG-vega | entry IV HIGH (above) | 33 | +9 | 76% | -6.7 | 1.1 | +7 | -2 | +4 |
| SHORT-vega / INVERSE | all weeks | 109 | -67 | 50% | -6.2 | -1.0 | -46 | -7 | -14 |
| SHORT-vega / INVERSE | entry IV LOW (below 90d median) | 76 | -54 | 51% | -6.2 | -1.0 | -36 | -9 | -9 |
| SHORT-vega / INVERSE | entry IV HIGH (above) | 33 | -13 | 48% | -6.2 | -0.8 | -10 | +1 | -4 |
| structure | IV regime | n | Total% | Win% | Worst% | Ret/DD | 2024 | 2025 | 2026 |
|---|---|---|---|---|---|---|---|---|---|
| LONG-vega | all weeks | 109 | +36 | 76% | -9.4 | 2.2 | +20 | +6 | +10 |
| LONG-vega | entry IV LOW (below 90d median) | 76 | +30 | 79% | -9.4 | 2.4 | +11 | +9 | +10 |
| LONG-vega | entry IV HIGH (above) | 33 | +6 | 70% | -7.8 | 0.6 | +8 | -2 | +0 |
| SHORT-vega / INVERSE | all weeks | 109 | -28 | 41% | -5.4 | -0.8 | -26 | +2 | -4 |
| SHORT-vega / INVERSE | entry IV LOW (below 90d median) | 76 | -31 | 36% | -5.4 | -0.8 | -20 | -6 | -5 |
| SHORT-vega / INVERSE | entry IV HIGH (above) | 33 | +3 | 55% | -3.5 | 0.6 | -5 | +8 | +0 |
run_cal2w_volregime.py. long=2 strikes OTM, near=1 strike OTM,
GRID $1000. iv_high = DVOL above its trailing-90d median at entry. M2 bias-corrected 2024+. Green row =
the winning cell (long-vega, low IV).