Giulio Occhionero on Why Crypto Breaks the Risk Models Built for Equities

What two decades of systematic equity work does, and does not, transfer to digital assets

Most of the quantitative apparatus we use to trade equities was built on assumptions that digital assets quietly violate. Returns are roughly normal in the middle and only occasionally extreme. Markets close, which gives risk a natural clock. Correlations are stable enough to build a portfolio around. None of these hold cleanly in crypto, and the practitioners who lose money there are usually the ones who assumed they would.

I trade systematic strategies across US equities, options, and crypto, and I have spent enough time on the digital-asset side to be wary of anyone who claims the same models port over unchanged. They do not. What transfers is the discipline. What has to be rebuilt is almost everything downstream of the distribution.

1. Fat tails are the base case, not the exception

In equity markets, a five-sigma day is an event you might see a handful of times in a career, and you build risk systems that treat it as rare. In crypto, moves that would be tail events for the S&P 500 arrive on ordinary weekends. The distribution is not lognormal with occasional shocks. It is heavy-tailed as a resting state, and the tails are asymmetric in ways that shift with the funding environment.

The practical consequence is that any risk model calibrated to a Gaussian core will chronically understate the probability of ruin. Value-at-risk computed the standard way is close to useless here. The models that survive are the ones that size positions off empirical tail behavior rather than off a fitted variance, and that assume the worst observed drawdown is a floor, not a ceiling.

If your risk model has never seen the move that ends your fund, it will price that move at zero right up until it happens.

2. A market that never closes has no natural clock

Equity risk management is quietly organized around the closing bell. Positions get marked, exposures get rebalanced, and the overnight gap is a known unknown you can hedge or reserve against. Crypto has no close. Liquidity thins on weekends and in the hours between the major regional sessions, and that is precisely when the largest dislocations occur.

A strategy that assumes it can always exit at a fair price is making an equity assumption in a market that does not honor it. The models that hold up treat time-of-day and day-of-week as first-class risk variables, reserve capacity for the thin hours, and never assume that the ability to trade at noon on a Tuesday says anything about the ability to trade at three in the morning during a cascade.

3. Correlation that betrays you at the worst moment

The diversification story in crypto is weaker than the marketing suggests. In calm conditions, tokens display enough dispersion to make a market-neutral book look attractive. Under stress, that dispersion collapses. Almost everything correlates to one macro factor, which is broad risk appetite expressed through the largest assets, and the carefully constructed neutral book turns out to be a concentrated directional bet at exactly the moment you needed it not to be.

This is the same failure that broke lazy risk-parity implementations in traditional markets, only faster and more violently. Modeling correlation as a constant is the error. Modeling it as a regime-dependent variable that trends toward one under stress is the beginning of a usable framework.

4. Liquidity that is deep until it is not

Order books in digital assets are fragmented across dozens of venues, and aggregate depth is misleading. The book looks thick because you are summing liquidity you cannot actually access simultaneously without moving every venue against yourself. Depth also evaporates nonlinearly. The first increment of size trades near the quote, and the next increment trades far worse, because market makers widen or pull the moment they sense a large participant.

Backtests that assume you can transact meaningful size at the mid are not modeling crypto. They are modeling a frictionless abstraction that has never existed. Realistic impact curves, venue-by-venue, are not a refinement here. They are the difference between a strategy and a story.

5. Reflexivity and the absence of an anchor

Equities have fundamentals to argue about. Cash flows, multiples, and balance sheets give price a tether, however loose. Much of crypto has no comparable anchor, which means price is driven more purely by positioning, narrative, and leverage. The market is reflexive: rising prices attract flows that justify higher prices, until the leverage that fueled the move becomes the mechanism of its reversal.

For a systematic trader, this argues for putting more weight on positioning and flow data, which are unusually observable in this market through funding rates, open interest, and on-chain movement, and less weight on any notion of fair value. The edge is in reading crowding and leverage, not in estimating an intrinsic price that may not exist.

6. Sizing when volatility is the only stable input

If the distribution is unstable, the correlations are regime-dependent, and the liquidity is fragmented, the one variable you can measure with reasonable confidence is realized volatility. That makes volatility-targeted sizing not just a preference but close to a necessity. Fixed notional sizing in an asset whose volatility can quadruple in a day is a way to be right on direction and still be liquidated on the path.

The strategies I have seen compound in this market share a conservative posture: they size small relative to what the backtest permits, they cut faster than feels comfortable, and they treat leverage as the primary killer rather than as a return enhancer. The asymmetry is unforgiving. You can be early, correct, and still gone.

What this comes down to

Crypto rewards the quantitative mindset and punishes the transfer of quantitative assumptions. The rigor, the skepticism toward tidy backtests, and the respect for capacity all carry over. The specific machinery, the Gaussian cores, the constant correlations, the frictionless fills, and the closing-bell risk clock, does not. The practitioners who do well here are the ones who kept the discipline and rebuilt the model. The ones who struggle are the ones who assumed a market this different could be traded with tools designed for a market this settled.

About the author

Giulio M. Occhionero is Head of Quantitative Research and Development at IRH Global Trading in Abu Dhabi, where he leads systematic strategy deployment and quantitative research across US equities, options, and digital assets. He holds degrees in nuclear engineering and has spent more than two decades in quantitative finance across roles in Italy, the United States, and the United Arab Emirates. He writes regularly on systematic trading, market microstructure, and the engineering side of investment management.