The decision is about collateral scope
| Question | Isolated margin | Cross margin |
|---|---|---|
| What backs the position? | Collateral assigned to that position/market | Eligible collateral shared within the margin account |
| Can another position affect it? | Usually separated, subject to venue/account rules | Yes; losses, gains, orders, and collateral haircuts can interact |
| Maximum immediate collateral exposure | Defined by the allocated margin plus any auto-add setting | Can extend to more eligible account equity |
| Operational risk | Underfunded position liquidates sooner | One position can consume buffer needed by others |
Bybit’s current documentation describes isolated margin as ring-fenced by position and cross margin as shared collateral. Binance similarly describes isolated margin as an independent margin account for each pair. Exact calculations, auto-margin settings, and liquidation steps remain exchange- and product-specific.
One price move, two failure paths
Suppose two traders open the same $10,000 perpetual position at the same price. Each initially assigns $1,000 of margin. Trader A uses isolated margin with auto-add disabled. Trader B uses cross margin in an account with another $4,000 of eligible equity.
The trade’s price P&L is identical. What differs is the buffer. Trader A’s position approaches its maintenance threshold using the isolated pool. Trader B can draw on more account equity, so the position may survive longer—but the loss can now consume capital reserved for other positions. Cross margin did not improve the trade; it widened the blast radius.
Stress the account, not only the position
| Shock | Isolated question | Cross question |
|---|---|---|
| Position loses 5% | Does assigned margin still clear maintenance? | How much shared equity remains for every position? |
| Funding debit | Does the debit reduce this position’s buffer? | Does it consume account collateral needed elsewhere? |
| Second position opens | Is its collateral separately assigned? | How does aggregate maintenance and order margin change? |
| Collateral asset falls | Is that asset inside the isolated pool? | Does a collateral haircut weaken the entire account? |
A decision example with two simultaneous positions
Assume a trader holds a directional BTC long and an ETH short intended as a hedge. Cross margin can be coherent if both positions are sized and reviewed as one portfolio and the trader accepts that either leg can consume shared collateral. The same cross setting is dangerous if the ETH short is an unrelated discretionary trade: its loss can erase the BTC trade’s planned buffer. Isolated margin makes those budgets explicit, but each leg must still carry enough collateral to survive normal volatility before its planned stop.
Same leverage does not mean the same liquidation path
Two positions can both display 10× leverage while one uses a fixed isolated allocation and the other sits inside a cross account with several collateral assets, open orders, and other positions. Their opening notional may match, but available equity, maintenance requirements, collateral haircuts, and fee/funding debits do not. Compare the live margin balance and maintenance requirement—not only the leverage label or a static liquidation calculator result.
Funding and fees can change the boundary
Funding and fees are cash flows against available margin. On some venues an isolated-position funding debit comes from the position margin; on cross it comes from cross-account equity. If the debit reduces the relevant pool below maintenance requirements, cancellation or liquidation logic can follow. Do not assume an eight-hour interval or one debit source across every contract.
A practical selection checklist
Consider isolated when
- you need a hard collateral boundary around an experimental position;
- the position should not consume buffer from unrelated trades;
- you can monitor the isolated margin and liquidation distance directly;
- auto-add margin is deliberately configured rather than left implicit.
Consider cross when
- positions are intentionally managed as one hedged account;
- you understand which assets are eligible collateral and their haircuts;
- the portfolio-level margin model is part of the strategy;
- you have account-level loss limits that are stricter than the exchange’s liquidation engine.
Journal the mode as risk data
For every leveraged trade record the venue, account mode, margin mode, contract family, side, size, leverage, entry, maintenance-margin tier, liquidation estimate, allocated collateral, funding settlements, and any margin added or removed. Then review whether losses came from the market thesis or from using the wrong collateral boundary.
Use the liquidation calculator for a scenario estimate, not as a replacement for the exchange’s live mark price, tier, fees, and account rules. Use the margin calculator to separate position notional from required margin.
Failure modes worth tagging
- Hidden portfolio coupling: an unrelated cross-margin loss consumed the planned buffer.
- Auto-add surprise: more collateral was assigned than the trade plan allowed.
- Tier jump: aggregate size moved maintenance margin to a higher bracket.
- Wrong invalidation order: liquidation sat inside the planned stop or thesis invalidation.
- Collateral mismatch: the collateral asset fell while the leveraged position lost.
These tags turn margin mode from a static setting into evidence. Review whether cross actually reduced forced exits for a hedged portfolio or merely allowed one bad position to remain open longer.
Run this margin-mode stress test before entry
- Write the planned stop, expected slippage, and the maximum account loss in currency terms.
- Record the exchange’s live initial margin, maintenance margin, liquidation estimate, and collateral assets.
- Shock the position to the planned stop and add the next funding debit plus a closing-fee estimate.
- For cross margin, shock the largest correlated position and any non-stable collateral at the same time.
- For isolated margin, test whether auto-add is enabled and whether the assigned margin survives without it.
- Reduce size or change the collateral boundary if liquidation can begin before strategy invalidation.
Metrics for the later review
| Metric | What it diagnoses |
|---|---|
| Planned stop distance vs liquidation distance | Whether the trade structure allowed the thesis to fail before the exchange intervened |
| Maximum shared collateral consumed | The true blast radius of cross margin |
| Margin added after entry | Whether isolated risk was actually capped or repeatedly rescued |
| Funding and closing fees as % of risk | Costs that silently narrowed the buffer |
| Other-position contribution to margin loss | Portfolio coupling rather than trade-level error |
Turn the choice into evidence
In TSB Journal, tag the exact margin mode and review outcomes by comparable setups. A useful test asks: did cross margin prevent a planned stop from doing its job, or did isolated margin force liquidation before the strategy’s invalidation? The answer must come from your positions and account rules, not a universal slogan.
