The thing that trips people up about margin is that the same book of positions can require wildly different amounts of collateral depending on which system is scoring it. I have watched a hedged pair that costs almost nothing under one framework eat tens of thousands of dollars in requirement under another, on the exact same positions, at the exact same prices. Nobody moved. The rules just disagreed about what the risk was. So if you are running anything more complicated than a handful of long stock positions, it pays to know which system your broker or exchange is actually using, because that choice quietly decides how much capital you get to deploy and how badly a single trade can hurt you.
Four ways to score the same risk
Reg-T is the old US default for stock accounts, and it is dumb on purpose. It looks at each position more or less in isolation and charges you a flat percentage, historically around fifty percent initial on a long stock position. It does not care that you are also short a correlated name that would offset the loss. Two positions that hedge each other still get charged twice. Reg-T is simple and predictable, and it is expensive if you actually hedge.
Portfolio margin is the smarter cousin, available to larger accounts that qualify. Instead of a flat rate per position, it shocks your whole book through a set of hypothetical price moves, typically stressing the underlying up and down by some band, and charges you based on the worst-case loss across the entire portfolio. A long position offset by a short in the same or a correlated name nets down, so your requirement can drop to a fraction of the Reg-T number. That is the whole appeal. You get more buying power for the same risk because the model finally recognizes your hedges.
SPAN is the futures and options world's version of the same idea, and it has been around for decades. It builds a risk array for each instrument, running a grid of price and volatility scenarios, then combines them across your book with credits for offsetting positions in related products. Calendar spreads and inter-product spreads get real relief. If you trade futures, your requirement is a SPAN number whether you think about it that way or not.
Crypto exchange unified accounts are the newest entry, and they borrow the logic without necessarily borrowing the rigor. A unified or portfolio margin mode on a large exchange lets your spot, perps, and options share one collateral pool and net against each other. A short perp can offset a long spot, options greeks fold into the same calculation, and your effective requirement drops. It feels like portfolio margin, and conceptually it is, but the stress parameters and the liquidation logic are the exchange's own, they can change with little notice, and they are tuned for assets that move far faster than equities.
Where netting actually helps
The reason any of this matters for a multi-asset book is that netting is where the capital comes from. Under a portfolio approach, offsetting exposures reduce your total requirement instead of stacking. A few places this shows up:
- A long spot position hedged with a short perp or future nets close to flat, so you hold a fraction of the collateral you would under a per-position charge.
- Options positions with opposing greeks partially cancel, so a defined-risk spread costs far less than the two legs charged separately.
- Correlated longs and shorts across related names get partial credit, so a relative-value trade is cheap to carry.
The trap is that all of these offsets are only as good as the correlation the model assumes. The model is pricing your hedge as if the two legs will keep moving together. In a normal tape they do. In a real dislocation, correlations that the risk engine treated as reliable can snap, both legs move against you at once, and the offset you were counting on evaporates right when you need it. Your requirement can jump hard in the same session the market is already hurting you.
Cross-margin is one shared fuse
Here is the part people learn the expensive way. When your positions share one collateral pool, they also share one liquidation trigger. That is what cross-margin means. Every position is backing every other position. So a single trade that blows out, some illiquid alt you sized too big, a short that goes parabolic, can burn through the shared equity and force the engine to liquidate positions that had nothing to do with the loss. Your carefully hedged, perfectly reasonable trades get sold to cover the one that went wrong. The blast radius is the whole account, not the bad position.
Isolated margin is the opposite. You wall off collateral per position, so a blowup can only lose what you allocated to it and the rest of the book survives. You give up the netting efficiency, and you pay for it in tied-up capital, but you cap the damage. The honest tradeoff is capital efficiency against blast radius, and you do not get both.
The way I think about structuring it is to sort positions by how much I trust them to behave under stress, then decide what shares a fuse with what:
- Put your core, liquid, genuinely hedged positions in a cross-margined pool where netting earns you real capital and the correlations are the kind that mostly hold.
- Push anything speculative, illiquid, or high-volatility into isolated margin or a separate sub-account, so its worst day cannot reach into the rest of the book.
- Size every isolated bet as if its liquidation price will be hit, because for the ones that matter, eventually one is.
- Track your margin requirement as a live number, not a number you checked when you opened the trade, since a volatility spike can reprice your whole book while you sleep.
- Keep a real buffer of unencumbered collateral, because the moment you need to add margin is exactly the moment everything is gapping and you cannot move fast enough.
That last one is the difference between a bad day and a wipe. Most liquidations I have seen up close were not caused by a wrong directional call. They were caused by running the account so tight that a routine repricing of the requirement left no room, and the engine closed positions at the worst possible prices to protect itself, not you.
Knowing the number before it moves
The practical skill is being able to answer, at any moment, what your requirement would be if volatility doubled and your hedges stopped hedging. If you cannot answer that quickly, you are carrying more risk than the current comfortable-looking requirement suggests, because the requirement is a function of market conditions and those change without asking you. On a multi-asset book that spans crypto and traditional markets, this gets harder because the same underlying can carry very different margin treatment on a stock desk versus an exchange unified account, and it is easy to double-count relief you only get in one place.
When I am watching how a book behaves across venues, having positions and exposure in one view helps, which is part of why we built the multi-asset side of Blockcircle the way we did. But the tool only tells you where you stand. The structuring decision, what shares a fuse with what, is yours, and it is worth making deliberately before the market makes it for you. Decide the blast radius while it is quiet. It is the one part of margin you actually control.