Crypto portfolio margin evaluates positions together under defined stress scenarios, allowing eligible gains on one leg to offset losses on another. The requirement depends on the venue's risk groups, scenario set and additional charges. A lower requirement does not cap possible losses: inspect which scenario dominates, whether the hedge remains present and how concentration changes the result.
Ask what is being offset
A portfolio can have large opposing positions and a small combined loss in one scenario. Portfolio margin recognizes some of that relationship when its rules allow it. The useful question is not simply whether two trades point in opposite directions, but whether the engine groups them together and values them under the same stress.
Bybit's Unified Trading Account documentation describes stress tests of underlying mark price and implied volatility within risk units. Its maintenance calculation combines maximum scenario loss with contingency components. Spot holdings require its Spot Hedging setting to participate in those tests; borrowing has a separate calculation. These are Bybit rules, not a universal portfolio-margin formula. Bybit's portfolio-margin calculation and eligibility rules.
The collateral boundary is another question. Use the cross-versus-isolated margin guide to identify which resources support positions. Portfolio scenario netting explains how a requirement is assessed; the account boundary explains which resources may meet it.
Sum each scenario before selecting the loss
Consider a hypothetical portfolio with one cash-funded BTC in spot and a short linear future representing one BTC. Both reference prices start at $50,000. Assume the two legs qualify for one risk group, all values use the same dollar unit and no currency conversion changes. Ignore borrowing, fees, funding, collateral haircuts and extra charges.
For this worksheet only:
spot price PnL = 1 BTC × (stressed spot price − $50,000)
short-future price PnL = 1 BTC × ($50,000 − stressed future mark)
This is an invented three-scenario diagnostic, not an exchange risk matrix or an adequate live stress set.
| Hypothetical scenario | Spot price | Future mark | Spot price PnL | Short-future price PnL | Combined price PnL |
|---|---|---|---|---|---|
| Shared decline | $40,000 | $40,000 | −$10,000 | +$10,000 | $0 |
| Shared rally | $60,000 | $60,000 | +$10,000 | −$10,000 | $0 |
| Future premium widens | $50,000 | $51,500 | $0 | −$1,500 | −$1,500 |
The largest loss among these combined observations is $1,500. Each leg separately has a $10,000 worst loss in the table, but those losses occur in different rows. Adding the two standalone worst losses gives $20,000 and discards the simultaneous scenario relationship. Looking only at the two shared moves makes the opposite error: it misses the $1,500 basis loss.
A compact expression for the worksheet is tested loss = max(0, −minimum combined scenario PnL). The zero floor prevents a set containing only gains from producing a negative loss measure. This convention describes the teaching calculation, not any venue's full margin formula.
The $1,500 is neither the cash needed to open the portfolio nor its liquidation threshold. An actual engine may evaluate more price moves, volatility changes, expiry effects, orders and additional charges. A single-position risk-limit tier calculation cannot replace that portfolio assessment.
A neutral snapshot can hide concentration
Scale the hypothetical holdings to four BTC spot and four BTC short. Keeping exactly the same linear assumptions multiplies each PnL by four: the shared-move rows still net to zero, while the premium-widening row loses $6,000.
The apparent neutrality of the first two rows survives the size increase. The absolute basis exposure does not. In a real account, increasing size can also change charges or the practicality of reducing both legs. Do not infer a fourfold margin requirement from the fourfold teaching loss.
Deribit's portfolio-margin documentation includes price and volatility scenarios, an extended table for large far-out-of-the-money short-option exposure, and additional delta and roll shocks. It also states that its Cross Portfolio Margin groups netting by base currency: a BTC short does not offset an ETH long for this purpose. Sharing collateral across currencies therefore does not establish cross-asset risk netting. Deribit's portfolio-margin risk matrix and grouping rules.
For options, a near-zero current delta is only a local price sensitivity. It does not establish that the whole portfolio stays neutral after a large move or volatility change. Read the stressed valuations of the actual contracts rather than replacing them with today's net delta.
Removing a hedge can raise the requirement
Return to the one-BTC portfolio. Suppose an operator sells its spot leg at the unchanged $50,000 reference and leaves the short future open. Under the same three scenarios, the future alone produces +$10,000, −$10,000 and −$1,500. The largest tested loss rises from $1,500 to $10,000: an $8,500 increase.
Selling an asset reduced gross exposure, yet removed the offset. That is the failure this worksheet is meant to expose. The result does not imply that the sale loses $8,500 immediately. It changes the subsequent scenario loss in this construction, while the sale proceeds and their collateral eligibility need separate accounting.
Test intermediate states before a multi-leg change: neither leg fills, only the spot fills, only the future fills, and both fill. Treat the actual filled quantity as authoritative. Two intended instructions do not establish one completed hedge.
Working orders can leave those states possible longer than expected. If an operator expects a hedge-changing order to expire, the good-till-date order guide explains the timestamp and execution boundary that must be verified. A local expiry clock is insufficient evidence that no late fill changed the portfolio.
Keep equity shocks separate from scenario requirements
The account can lose support even when its position quantities are unchanged. A funding payment changes booked cash; collateral repricing can change recognized equity. Neither should silently disappear inside the scenario-PnL column.
For example, attach an invented $200 funding debit to the unchanged one-BTC portfolio. The three price-PnL rows remain as written, while cash equity falls by $200. This is a separate ledger adjustment, not evidence that a real venue's requirement remains fixed. The funding-timestamp position-snapshot guide helps establish which holding and payment belong to the account observation.
Likewise, apply the stablecoin collateral stress worksheet to the support side. Record valuation currency, eligibility and haircut assumptions before comparing recognized equity with requirements. Use the venue's dated account metric to assess its actual state; the mark-price and liquidation guide explains why an ordinary trade-price chart cannot substitute for that evidence.
Save a scenario dependency record
| Field | Evidence to retain | Question it answers |
|---|---|---|
| Account and model | Venue, account mode, valuation currency, observation time | Which calculation applies? |
| Eligible grouping | Exact contracts, base currencies and eligible spot holdings | Which legs may offset? |
| Scenario inputs | Price moves, volatility assumptions and parameter revision | What was actually tested? |
| Dominant loss | Combined PnL by scenario and contributing legs | Which dependency drives the result? |
| Extra requirements | Applicable concentration, basis, expiry and order charges | What lies beyond the headline scenario loss? |
| Transition states | Current fills and each possible one-leg result | Does the offset survive execution? |
| Supporting equity | Collateral valuation and booked cash flows at the same time | What resources meet the requirement? |
Keep the original matrix with this record. A later screenshot showing a smaller requirement cannot explain whether the portfolio improved, a hedge appeared, or the venue changed parameters.
Finally, “worst” refers to the tested set. Deribit explicitly warns that its worst-case scenario label does not mean the greatest amount an account could lose. Deribit's explanation of the tested-scenario boundary. Market moves outside the grid, disrupted offsets and difficult exits remain possible. The record makes those dependencies visible without turning a model output into a loss guarantee.
Editorial method: Prepared with AI assistance on 23 October 2026 from the official sources linked above. The three-scenario portfolio, concentration extension and hedge-removal calculations are original hypothetical examples. No live account, exchange margin engine or KLINEO portfolio-margin feature was tested.