The Structural Short-Vol Risk

Lesson 2 sold you the dream: drop ETH into a covered-call vault, the protocol auctions a call every Thursday, and the premium drips back to you as a juicy double-digit “APY.” Free money, denominated in the asset you already wanted to hold. What’s not to love?

This is the lesson where we read the label on the bottle. That APY is not interest, not staking reward, and not a fee the protocol generously shares. It is an insurance premium you collect for taking on a very specific, very lumpy risk. A DeFi option vault (DOV) is, structurally and unavoidably, a volatility seller — and selling volatility has a payoff shape that looks gorgeous right up until the week it doesn’t.

A vault is a volatility seller in disguise

Strip away the branding. Every Thursday the vault writes an option — a call (covered-call vault) or a put (cash-secured-put vault) — and sells it at auction. Whoever buys it is paying for the right to profit from a big move. The vault, by writing it, has promised to eat that big move. In options language, the vault is:

• Short the option — it sold something with open-ended risk.
• Short vega — its position loses value when implied volatility rises (when the market gets scared and option prices balloon, the thing the vault already sold gets more expensive to buy back).
• Short gamma — as the underlying moves, the vault’s delta moves against it. Up moves hurt a short call more and more; down moves hurt a short put more and more. Short gamma means your losses accelerate exactly when you’d want them to slow down.

So what is the “yield”? It’s the volatility risk premium (VRP) — the same beast you met in the Volatility Trading course. Option prices embed an implied volatility that, on average, sits above the volatility that actually shows up (realized). Buyers systematically overpay for protection because they hate getting wiped out, and that fear has a price. The seller pockets the gap.

Vault APY ≈ harvested VRP = (implied volatility the buyer paid) − (realized volatility that materialized), converted into premium income.

That’s it. The vault is a machine for renting out your willingness to absorb tail risk. The double-digit APY is the rent. And rent on a fire-insurance policy looks fantastic in every year there’s no fire.

Negative skew: pennies in front of a steamroller

Here is the shape of the bet. Most weeks, the option the vault sold expires worthless or nearly so, and the vault keeps the whole premium. Small win. Small win. Small win. Then — rarely, unpredictably — the market gaps, the option goes deep in the money, and the vault hands back far more than every premium it ever collected. One big loss.

Many small wins, one occasional large loss. That is a negatively-skewed payoff, and it is the textbook profile of “picking up pennies in front of a steamroller.” The win rate is flattering. The expectancy is whatever’s left after the steamroller takes its turn.

The chart below is the VRP made visual: implied vol (what the vault sells) sits a few points above realized (what actually happens) through the calm stretches — that green band is the rent. Then watch the crisis episode, where realized vol gaps above implied and the seller eats a sudden, large loss (the red band).

Worked example: 50 quiet weeks, one ugly one

Let’s price the steamroller. Suppose a covered-call vault collects a clean +0.4% premium per week on the deposited notional, and it does this for 50 weeks straight without a hitch. Then, in week 51, ETH gaps −25% in a single weekly cycle and the option the vault sold blows up.

The arithmetic: 50 × 0.4% = +20.0% of harvested premium. Looks like a 20%+ annualized machine. Then one −25% week: 20.0% − 25.0% = −5.0% net over the whole year.

A 50-out-of-51 win rate (98%!) and the strategy is down on the year. That’s negative skew in one line. The premium income is real, steady, and seductive; it is also, in expectation, compensation for that −25% week — not a free lunch on top of it. If the crash week pays out more than ~50 quiet weeks can store up, you can have a glittering hit-rate and still bleed.

Why the auctions can underprice the options

Now the insult on top of the injury. Even if you want to be short vol — fine, harvesting VRP is a legitimate strategy — you’d at least like to be paid fairly for it. Vault auctions frequently don’t deliver fair value. They can clear below the option’s theoretical price, which means depositors are short vol and getting a bad fill.

Why does the price leak?

• Thin bidder sets. A weekly DOV auction might attract a handful of professional market makers, not a deep competitive order book. Few bidders, lower clearing price. The vault is a price-taker selling into a near-monopsony.
• Predictable, telegraphed flow. Everyone knows the vault sells the same kind of option, roughly the same size, on the same day, at a strike chosen by a known rule. Predictable flow is a gift to the other side: market makers can pre-position, shade their bids, and let the vault come to them. You don’t get the best price when your counterparty has your calendar.
• Size pressure. A large vault must offload a lot of notional in one shot. Dumping size into a shallow auction moves the price against you — the same slippage you’d pay market-selling any asset, now applied to your premium.

Put the two ideas together. Fair VRP harvesting is a thin edge to begin with — implied minus realized is a few vol points. If the auction shaves another chunk off the premium through underpricing, the depositor is left holding the full negative skew of the short option while collecting only a fraction of the premium that risk deserves. You kept all of the steamroller and sold half the pennies.

Crowding and reflexivity

It gets worse when the trade gets popular — and short-vol trades always get popular, because the equity curve looks irresistible right up until it doesn’t.

When billions of dollars of vaults all sell the same strikes on the same weekly cadence, the dealers buying those options end up holding a giant, homogeneous book of the same gamma. To stay delta-neutral, every one of those dealers has to hedge the same way at the same time. Short-gamma-from-the-vault means the dealers are long gamma, so the dealers buy as the market falls and sell as it rises — usually stabilizing. But the positioning is lopsided and concentrated, and that creates two distinct hazards.

• Pinning / max-pain near popular strikes. As expiry approaches, dealer hedging tends to drag the spot toward the strike where the most open interest sits (the “max pain” point), because that’s where their hedging flows net out. Price gets magnetized to the crowded strike. Harmless-looking — until the magnet snaps.
• Reflexive unwinds. If the move is large enough to blow through the crowded strikes, the hedging flips. Now everyone is offside on the same position, and the de-risking is one-directional: the vaults’ losses, the dealers’ re-hedging, and forced unwinds all push the same way at once. A feedback loop. The crowd that smoothed the small moves amplifies the big one.

The core lesson: crowded short-vol is more dangerous than lonely short-vol. When you’re the only one selling insurance, your loss in a crash is your own. When the whole market is short the same vol on the same schedule, your crash is everyone’s crash, all the exits are the same door, and the tail you’re short gets fatter precisely because so many people sold it.

What a crash does to “yield”

Time to watch the steamroller actually drive over the vault. Take a real drawdown and trace each vault type.

Covered-call vault in a crash. The vault holds the asset and sold an upside call. In a crash, the upside call expires worthless — so the vault keeps that one week’s premium, maybe +0.4%. Cold comfort: the vault still holds the asset, which just fell 25%. The premium is a thin cushion taped to the front of a falling object. Worse, by writing the call the vault capped its upside on the eventual recovery, so it eats the full crash but clips the rebound. You absorbed the whole left tail and rented out the right tail.

Cash-secured-put vault in a crash. The vault posted stablecoins and sold a downside put. In a crash, that put goes deep in the money and the vault is assigned — it’s forced to buy the falling asset at the (now far-above-market) strike. It collected, say, +0.4% in premium and is now sitting on an asset that’s down 25% from the strike it paid. The premium covered roughly 1/60th of the loss.

Either way, the annualized APY that the marketing page advertised can be erased by a single event. Recall the worked example: a 98% win rate and a −5% year. The APY box never lied about the calm weeks; it just couldn’t show you the one week that matters.

Now the most important reframe. The chart below contrasts the two payoff shapes. The vault is the concave seller — capped gain, open-ended loss, the steamroller profile. The person on the other side of the auction is the convex buyer — small known cost, large open-ended upside. Toggle between the shapes:

And here is what that buyer’s position actually does to a portfolio in a crash — the mirror trade. A strip of deep out-of-the-money puts costs a small, steady premium drag in calm markets (exactly the premium your vault is collecting), but it floors the loss and even turns convex in a true tail:

Stare at those two charts together. The vault is selling the very convexity the tail-hedge buyer wants. Your weekly premium is their premium drag. Their crash payoff is your crash loss. If you’ve internalized that the smart, defensive move in the Volatility Trading course was to buy cheap tail protection, then notice: the DOV puts you on the opposite side of that trade, every single week. That is not automatically wrong — someone has to sell the insurance, and the VRP is a real edge — but you should know which seat you’re sitting in.