Short Puts in a Market-Wide Drawdown: 2018, 2020 and 2025

The risk that does not show up in a single position

Analyse one short put and you find a bounded, dated obligation with a known worst case. Analyse twelve of them and you find something different, because in the conditions that matter they stop behaving like twelve independent positions and start behaving like one.

Three episodes in recent memory show what that looks like. They differ in cause, duration and severity, and they share a structure worth understanding before the next one.

February 2018: the volatility spike with no crash attached

The remarkable thing about Volmageddon is how little happened to the stock market.

On Monday 5 February 2018 the VIX closed at 37.32, up 115.6% from its close of 17.31 the previous Friday — the largest single-day percentage rise in the index's history. The S&P 500 fell about 4% that day. Unpleasant, unremarkable, the kind of decline that occurs in most years.

The volatility complex did not treat it as unremarkable. XIV, an exchange-traded note that was short VIX futures, lost more than 90% of its indicative value in a single after-hours session and finished the day down roughly 97%. Credit Suisse announced its liquidation the following day.

For a put seller the lesson is in that gap between a 4% equity move and a doubling of implied volatility. Short puts are short volatility, and the volatility component repriced far more violently than the underlying did. A book of open short puts would have shown mark-to-market losses out of all proportion to the move in the stocks themselves, because vega repriced before delta had done much of anything.

Positions that were never going to be assigned still looked alarming, and traders who reacted to the marks rather than to the strike distances closed at the worst prices of the episode. The event resolved quickly. The behaviour it triggered did not always resolve with it.

March 2020: the one where everything moved together

COVID was the full version. The S&P 500 peaked at 3,386.15 on 19 February 2020 and bottomed on 23 March, a decline of roughly 34% in about five weeks. The VIX closed at 82.69 on 16 March, its highest close on record.

Three things happened to short-put books simultaneously.

Diversification stopped functioning. Correlations across equities converged toward one. A book spread across ten names in six sectors behaved like a single leveraged position on the index, because in a liquidity-driven decline the market sells what it can rather than what it should.

Every strike was breached, in sequence. A 34% index decline over five weeks means the strike that was comfortable in week one was in the money in week three and deep in the money in week five. There was no single gap; there was a grind that consumed every buffer in the book one expiry at a time.

Rolling stopped being available on acceptable terms. The instinct when a short put goes against you is to roll down and out. In March 2020 the cost of doing so was prohibitive precisely because implied volatility was at a record, and the further-dated contracts a seller wanted to buy back into were the most expensive they had ever been. The escape route repriced at the same moment it was needed.

The recovery was unusually fast — the index made a new high within six months — which has since done a certain amount of damage. It taught a generation of sellers that drawdowns resolve quickly. The 2000 and 2007 analogues, which took years, teach something else.

April 2025: the fast one

The most recent episode was compressed into days.

Following the tariff announcements of early April 2025, the S&P 500 fell 4.8% on 3 April and a further 6.0% on 4 April. At its worst the index was down roughly 19% from its February record high. The VIX peaked at 52.33 on 8 April, and one-month observed volatility reached nearly 43%, the highest since 2020.

Then it reversed. A 90-day pause on implementation produced a 9.5% single-day rally in the index, among the ten largest daily moves in its history. The month closed down 0.7%.

That round trip is the instructive part. A short-put book that was closed in panic on 7 April recorded losses that the book would not have recorded had it been left alone, because most of those strikes were never breached at expiration. Equally, a book that survived did so partly on a policy reversal that nobody could have forecast, which is not a risk-management strategy.

The honest reading of April 2025 is that both the panicked seller and the steadfast one were exposed to something outside their control. One was rescued and one was not, and the difference was not skill.

What the three have in common

Correlation appears exactly when it is most expensive. In ordinary conditions a book across ten underlyings in six sectors carries ten roughly independent risks. In each of these episodes it carried approximately one. The diversification was real on the days it did not matter and absent on the days it did.

Implied volatility rises as positions deteriorate. For a seller this is doubly punishing: the mark-to-market loss on existing positions widens because vega is against you, and the cost of any defensive adjustment rises at the same time and for the same reason. There is no moment during the event at which repair is cheap.

New premium looks extraordinary and is not free. The most tempting screens a put seller will ever see appeared in the middle of all three episodes. Premium was at multiples of normal. It was also correctly priced for the conditions, and sellers who scaled up into it were adding exposure to the exact factor already hurting them.

The behavioural failure costs more than the market does. In two of the three, holding to expiry would have produced a materially better outcome than reacting to the marks. That is not an argument for never adjusting. It is an argument for deciding in advance what would cause you to adjust, because the decision made at the bottom is reliably the worst one available.

What actually reduces the exposure

Only two things do, and both are set before the event.

Correlated size, not nominal size. The relevant measure is not how much capital each position uses but how much of the book responds to one factor. Eight positions in different tickers that all fall when the index falls is one position eight times over. This is the subject of sector concentration, where the same effect is visible in a single day of settled contracts.

A pre-committed policy. Written down, before the drawdown, specifying what you will do at what level and what you will not do. Its value is not that the policy is optimal. It is that it exists at the moment when your judgement is least reliable, and the alternative to a mediocre plan is not a good plan but an improvised one.

Notably absent from that list: stops, which do not survive a gap; and rolling, which is available in ordinary conditions and repriced against you in exactly these.

How StratosIQ handles it

The model does not forecast regimes and does not claim to. What it does is refuse to relax the safety floor when premium becomes abundant, which is the specific failure mode all three episodes reward.

The floor comes from ShieldIQ, measured against how each underlying has actually behaved rather than against a model-implied distance, and reported as Fortified, Secure, Tight or Exposed. A contract failing that floor does not publish however rich it looks — which means that in the conditions where premium is at its most tempting, fewer contracts publish rather than more. Patrol's calibration is regime-aware, so the scoring adjusts to conditions rather than applying a fixed standard across every environment.

There is also a ceiling. Contracts whose buffer is unusually wide do not publish either, on both engines, because measurement showed that band carried a materially higher loss rate than the overall baseline. A buffer can be too wide as well as too narrow, and both ends are gates rather than preferences.

The full published record, contract by contract and including every loss, is on the performance page.

Worth reading next

Gap risk covers the single-name version of the same problem. Why a 90% win rate can still lose money explains why the rare, deep outcomes described here dominate the arithmetic of the strategy.