Module 6
Digital assets — what actually breaks
Every methodology in this track was designed for markets with a closing bell. Crypto has none — and most of what breaks in digital-asset margin traces back to that, plus custody.
- 24/7 markets.No end-of-day, no settlement cutoff, no overnight-MPOR convention. Margin must be computed and called continuously; the “nightly batch” concept dissolves. The engineering consequence is continuous incremental recompute rather than a nightly grid run (Module 7).
- Volatility scale. A ±15% equity scan range is a rounding error in crypto. Ranges run several times wider and must be dynamic— calibrated on rolling realized vol with a floor, so the range follows vol up quickly but won't follow it all the way down in a calm spell.
- Venue and custody risk. Collateral held at an exchange may be unreachable during a liquidation — the exact scenario in which you need it. Hence add-ons for venue concentration and haircuts differentiated by where the asset sits: tri-party custody versus exchange omnibus accounts.
- Stablecoin collateral. Near-zero market risk, non-zero depeg and redemption risk. The haircut is a credit judgment, not a vol calculation — no lookback window on a flat price series will tell you what happens when the peg breaks.
- The weekend gap. The margin period of risk is longer than it looks when fiat settlement rails are closed while crypto markets are not. A Friday-evening default may not be curable in cash until Monday — two extra days of a market that never stops moving.
🎛 Crypto scan-range calibrator
Calibrate a crypto scan range, then see what the collateral is really worth
Scan range on a $100,000 position
k·σ·√t is binding: 3.5σ over 4 effective days.
Margin required
$26,458
$120,000 stablecoin is worth
$111,720
Excess collateral
$85,262
Crypto breaks the equity conventions: vol runs several times higher, so the scan range is calibrated on rolling realized vol with a floor; markets trade 24/7 while fiat rails close on weekends, so the margin period of risk is longer than it looks; and collateral value depends on where it sits — an exchange omnibus balance may be unreachable in exactly the scenario where you need to liquidate. The stablecoin haircut is a credit judgment (depeg and redemption risk), not a vol calculation. Educational tool — not investment advice.
The procyclicality trap
Dynamic calibration has a dark side: if the scan range mechanically tracks realized vol, then a selloff raises margin, which forces deleveraging, which deepens the selloff — the margin model amplifies the crisis it is measuring. That is why the floor matters (keep margin honest in calm markets so it has less distance to jump) and why recalibration frequency is a policy choice, not just a statistics choice. Regulators worry about exactly this in CCP margin; a house model faces it in miniature.
Things to try
- • Calm regime (vol 30%): the floor binds. Crisis regime (vol 120%): k·σ·√t takes over and the range dwarfs the equity convention.
- • Toggle the weekend: two extra effective days move the range materially — √t is gentle, but not free.
- • Same book, flip custody from tri-party to exchange omnibus: the margin didn't change, but the collateral covering it did. Watch the excess flip to a margin call.
Test yourself
How would you set the haircut on a stablecoin posted as collateral? Why is “its 30-day realized vol” the wrong starting point, and what would you look at instead? (Reserve quality, redemption mechanics, concentration of the arb desks that maintain the peg — a credit analysis, not a market-risk one.)