Volatility harvesting

The volatility risk premium

Options markets usually price implied volatility above the realised volatility that follows. Sellers of options collect this difference as premium over time. This systematic gap — positive in most markets, most of the time — is what volatility harvesting captures.

The premium exists because buyers value the insurance options provide. They pay above actuarially fair prices to hedge their portfolios. Systematic sellers provide that insurance and collect the excess over time.

What the premium pays for

Mostly crash risk. Bollerslev and Todorov estimate that more than half of the S&P 500 variance risk premium is compensation for jump tail risk. In their estimates the actual distribution of returns had roughly symmetric tails, yet option prices weighted the left tail several times more heavily than the right. Carr and Wu found that market beta explains only a small part of the index premium, and the Fama-French factors do not explain it either. Sellers are paid mainly for holding the jumps that a delta hedge cannot follow.

Hedged buyers lose on average. Bakshi and Kapadia studied S&P 500 index options bought and delta-hedged with the underlying. The hedged positions lost money on average, and the shortfall was larger when volatility was high. The result held after allowing for fear of jumps, which points to a separate, negative price for volatility risk itself. This is the same premium seen from the buyer's side.

The premium lives at the short end. Dew-Becker, Giglio, Le and Rodriguez studied variance swap prices from 1996 to 2014. Hedging news about future variance, at horizons from one month to 14 years, was on average costless. Only unexpected, transitory realised variance carried a price. On this evidence, selling long-dated volatility adds exposure to volatility news that the market has not, on average, paid for.

Lessons from the equity record

1987 taught the market to price the left tail. Jackwerth shows that before the October 1987 crash, the S&P 500's option-implied and realised return distributions both looked roughly lognormal. Afterwards the realised distribution looked much the same, while the option-implied distribution turned left-skewed. Jackwerth and Rubinstein found that post-crash option prices made a three-standard-deviation annual fall about ten times more likely than lognormality implies. Index options have carried that put skew ever since.

Variance swaps failed when they were needed. Ian Martin records that the price jumps of 2008 and 2009 caused the single-name variance swap market to dry up completely. He traces this to the assumption behind the standard pricing and hedging theory that prices move without jumps. Quoting Carr and Lee, he notes that sharp moves exposed dealers to cubed and higher-order returns, and that they could not take positions in the deep out-of-the-money options the hedge required. Capping variance swap payoffs then became the market convention.

On 5 February 2018 the products amplified the fall. That day the S&P 500 fell about 4% and the VIX rose 20 points, its largest one-day rise since the 1987 crash. The BIS put the assets of select leveraged and inverse volatility products at about $4 billion at the end of 2017, and both kinds had to buy VIX futures just before their net asset values were struck at 16:15. The BIS saw signs that other traders began bidding up VIX futures from about 15:30 in anticipation. At 16:08, 115,862 contracts traded in a single minute, roughly a quarter of the VIX futures market, and the inverse product XIV lost 84% of its value before it was terminated.

The lesson is about predictable flow. Augustin, Cheng and Van den Bergen describe a feedback loop in which higher futures prices cut the products' assets and forced more buying, which lifted futures again. The BIS concluded that synthetic leveraged structures can create and amplify market jumps even when the core players are relatively small. VIX futures dealers hedged by selling S&P 500 futures, which added downward pressure on equities. The risk sat in a predictable daily rebalancing rule that other traders could see coming.

Implementation in crypto

Deribit is the dominant venue for crypto options. The market is deep in BTC and ETH options across a range of strikes and expiries. Implied volatility in crypto has historically run substantially above realised volatility — the premium is larger than in most traditional asset classes in volatility points, though not as a share of what the options cost.

A systematic strategy on Deribit sells options across a range of strikes, manages delta exposure daily, and rolls positions as expiry approaches. The return comes from the gap between what the market implies will happen and what actually occurs.

What is different in crypto

Bigger in level, smaller in share. Almeida, Grith, Miftachov and Wang studied Deribit bitcoin options from July 2017 to December 2022. They found a variance risk premium much larger than the S&P 500's in absolute terms. Measured against the level of implied variance, it was a smaller share than in Bollerslev, Tauchen and Zhou's S&P 500 estimates for 1990 to 2007. Most of the variance priced into bitcoin options in their sample was matched by the variance that followed.

The premium sits on both tails. Almeida and co-authors found the bitcoin premium larger in calm regimes than in volatile ones. In the calm regime the right side of the smile was steeper, and investors appeared more concerned about variance and upside risk. Deribit's index documentation notes that bitcoin options often show positive skew on both calls and puts, so a short call faces the kind of jump that a short put faces in equities.

The skew can change sign within a year. Block Scholes, writing on Deribit Insights, found bitcoin smiles skewed towards calls at all tenors on 7 January 2025. By 25 December 2025 the same weekly report found a put premium at all tenors, and CoinDesk reported that puts had traded at a premium to calls for most of the year. Deribit's index data show bitcoin's DVOL near 60 at the start of 2025 and near 43 at the end, after a low of about 33 in mid-September. The head of Wintermute's OTC desk, quoted by CoinDesk, described call overwriting against ETF and treasury holdings as the dominant flow of 2025.

Structural sellers move the price. On-chain option vaults sold calls and puts in weekly auctions, mostly on Fridays, because Deribit's Friday expiries were the most liquid. Analysis by OrBit Markets, published by Paradigm in September 2022, found that one-week ETH at-the-money implied volatility around Friday auction times averaged about 4 vols below the previous 24 hours. Traders learnt to anticipate the flow, so implied volatility tended to fall before each auction and the vaults sold into the dip. Institutions that wrote calls against spot and ETF holdings in 2025 added steady call supply of a similar kind.

Regime and EVAR

Volatility harvesting is the strategy most directly affected by regime. In a calm environment, implied volatility is moderate and realised is low — the premium is usually positive. In a spike, realised volatility jumps above the implied level that was sold, short positions lose, and positions require careful management.

EVAR for the volatility book uses a separate regime signal. The volatility regime is tracked independently of the broader market regime — a crypto volatility event does not always coincide with a macro regime shift. When the signal deteriorates, the book reduces before the event fully plays out.

What the price of volatility signals

Thin premiums have come before higher risk. Cheng finds that the premium embedded in VIX futures falls or stays flat when measured risk rises, and that falling premiums have predicted rising market risk. He links this partly to falling hedging demand, with premiums and trader exposures dropping together as risk rises. His measure is the premium in VIX futures, which price future implied volatility. Carrying the result over to the gap between implied and realised volatility in options is an inference.

Equity relationships may not carry over. Bollerslev, Tauchen and Zhou show that a high variance premium in the S&P 500 has predicted higher equity returns, most strongly at a quarterly horizon. In bitcoin, Almeida and co-authors compared two regimes and found the opposite pattern. The calm regime had the higher variance premium and the lower return premium. Their test compares regime averages, which is weaker evidence than a predictive regression, so a signal calibrated on equity data may not hold for bitcoin.

Know what the index leaves out. Deribit launched DVOL on 31 March 2021 as a 30-day measure built with variance swap methodology from the expiries either side of 30 days. It discards in-the-money options and far out-of-the-money options with delta below 5%, and publishes a smoothed average of a value computed every second. The far wings, which pay off only on the largest moves, therefore sit outside the index. DVOL measures only the implied side, so any premium estimate depends on the realised forecast set against it.

Events show up as kinks. Before the SEC's spot bitcoin ETF decision, Block Scholes tracked a volatility premium at expiries on or after the end-of-January contract, which put a kink in the at-the-money term structure of both bitcoin and ether. The kink moved along the curve as the decision date approached. The SEC approved the products on 10 January 2024.

Weekends still count. Crypto trades every day, and Deribit converts DVOL to an expected daily move using the square root of 365. Kaiko reported in June 2024 that the weekend share of bitcoin trading volume had fallen to a record low of 16% that year, from a peak of 28% in 2019. Weekend time decay therefore accrues while spot markets are thinner.

Where it breaks

Tail events. Options are non-linear. A sufficiently large move can overwhelm premium collected over months. Position sizing that accounts for this non-linearity — not just EVAR but specific strike selection — is how this is managed.

Volatility of volatility. Even in a generally calm market, implied volatility can spike and reverse quickly. A short volatility book has to manage intra-regime spikes without exiting prematurely on transient moves.

Correlation with other strategies. In a market stress event, volatility harvesting losses often coincide with losses across other strategy types. Correlation rises exactly when diversification is most needed. Portfolio-level EVAR accounts for this: when multiple strategies approach their individual EVAR ceilings simultaneously, the overall book reduces further.

Traps that are easy to miss

Variance and volatility differ. A variance swap is hedged with out-of-the-money options held in proportion to one over the strike squared, so low-strike puts carry the most weight in the hedge. Carr and Wu show that the variance swap rate exceeds the square of at-the-money implied volatility, and use the gap as a measure of the risk-neutral variance of volatility itself. Several of the studies cited here measure the premium in variance terms. The at-the-money gap is a narrower quantity, and a seller of variance is also short the wings and the volatility of volatility.

Coin-margined puts compound. Alexander and Imeraj noted in 2021 that most crypto options trading had been on Deribit's inverse contracts, whose dollar payoffs are margined and settled in BTC or ETH. They show that the delta of an inverse put grows without bound as the price falls towards zero, and that a deep in-the-money inverse put carries far more vega than an equivalent standard put. A short put margined in coin loses on the option while its collateral falls in dollar terms, and an education article on Deribit Insights warns that naked puts backed by BTC can lose very large amounts of BTC. Deribit's USDC-settled BTC and ETH options, announced for 19 August 2025, remove this compounding when the collateral is also stablecoin, though the short put still loses as the price falls.

Overwriting carries an uncompensated exposure. Israelov and Nielsen split a covered call into passive equity, short volatility and an active equity exposure that behaves like a reversal strategy. The active piece contributes significantly to risk and is not compensated. Vaults and overwriting programmes that leave delta unmanaged carry it. Israelov and Nielsen propose hedging it away, which leaves the equity and short-volatility exposures in place.