If you sell volatility systematically, you’ve been asking yourself one question for about six weeks now: when do I turn it back on?
A quick recap for anyone newer to this: the volatility risk premium is the gap between what options charge for future volatility and what the market actually delivers. Selling it is essentially running an insurance company—you collect premium most months and occasionally take a severe payout. Late February through March was the severe payout. If you were short vol going into the war, you took the hit the premium exists to compensate you for. Annoying, but that’s the contract you signed.
Here is the catch: the months following a major spike are historically the best months this strategy ever sees. Implied volatility has a long memory. Hedgers who just got burned keep overpaying for protection, dealers who ate losses charge more to write new paper, and every risk committee on earth suddenly approves buying insurance right after the fire. Realized volatility, however, doesn’t care about trauma; it inevitably subsides. That means the spread you are harvesting sits at its widest exactly now, in the decay phase, while everyone is still nursing their nerves.
This creates an ugly dilemma. Restart too early and you catch an aftershock—second spikes off elevated bases happen all the time. Restart too late and you sleep through the exact vintage that pays for the entire strategy.
My gut is useless here for the same reason it was useless at the March lows: emotional timing always arrives on the wrong schedule. The solution is the same as March—a checklist written in advance.
My checklist relies on five signals. I want most of them green, not all of them, and definitely not just one:
- Term structure normalized and holding: Front-month vol below longer-dated vol, establishing the ordinary upward slope. But the curve shape isn’t the signal by itself—any relief rally can fix the curve for a day. What matters is whether it holds through a bad headline. If the curve re-inverts every time Iran talks wobble, the market is still pricing emergency, and I am not selling fire insurance while the building is still smoking.
- Realized vol actually declining: Ten- and twenty-day realized volatility must sit clearly below implied. This rule makes me late on purpose. I will never capture the top tick of panic premium, and I am fine with that. The spread in the decay phase is wide enough that showing up deliberately late costs a little yield while skipping the worst of the aftershock window. That trade-off is the core philosophy.
- Vol-of-vol calming down: The market’s pricing of volatility on volatility is the best read on whether participants still fear another regime break or are returning to normal noise. If vol-of-vol remains jumpy, the market expects another gap, which is the exact environment where short-vol positions die. I want it boring before deploying real size.
- No scheduled binaries on the calendar: This is the April lesson. The ceasefire deadline never cleanly resolved; it smeared into rolling negotiations, vague proximity headlines, and policy caveats. A smeared binary is arguably worse for vol sellers than a clean one. A clean date can be stepped around; a smear is a constant drizzle of headline risk with no schedule. Rule: no short vol spanning an identifiable event date, and smaller size while the drizzle lasts. The premium will still be there when it resolves. My only real job is staying solvent long enough to collect it.
- Correlation back to normal: The same signals from the dispersion framework apply here—implied correlation declining, stocks trading on their own idiosyncratic stories again. Vol selling and dispersion are structural cousins, sharing a short exposure to the “everything moves at once” state.
Where I score it today: Items 1, 2, and 5 look mostly clear. Item 3 is getting there. Item 4 is the holdout—the negotiation drizzle is very much still falling.
That points to a specific action rather than a mood: restart at partial size, utilizing structures that cap damage if item 4 bites.
Execution and Safeguards
- On Sizing and Structure: Deploy in tranches—pre-planned steps as the signals season, never one massive allocation. Early tranches belong in defined-risk structures (spreads and condors) rather than naked strangles. You accept lower premium per unit for a known worst case, earning your way into fuller expressions later.
- The Convexity Budget: Spend a slice of the harvest on cheap, far out-of-the-money protection. It drags on returns every quiet quarter, but it is non-negotiable because the strategy is a bet on surviving long enough for the statistical average to play out.
- Pre-Commit the Shutoff Rules: Establish realized volatility thresholds, drawdown limits, and automatic pauses when new scheduled binaries appear. If you don’t pre-commit the off-switch, the decision gets made mid-panic by whoever in the room is most terrified.
Two final caveats: First, the post-spike vintage is the best on average, but averages hide path-dependence. Most decay phases pay beautifully; occasionally, one reruns the crisis. The checklist shifts the odds, but nothing deletes the tail. Second, everyone knows this trade. Post-crisis premium is famous, and famous premium gets crowded, which can compress the harvest faster than history suggests.
Insurance companies aren’t heroic; they are procedural. They survive by maintaining underwriting rules that don’t require anyone to be brave at the worst possible moment. That is the difference between a trade and a program—a trade requires me to be right about a specific restart, while a program requires the rules to be right on average.
The machine goes back on this week. Partial size, defined risk, per the list. Not because it feels safe—it doesn’t—but because the list says so, and the list was written by a calmer version of me.