Cross Margin vs Isolated Margin on Hyperliquid Perps: When to Use Each

If you are running more than one perp position on Hyperliquid, the choice between cross margin vs isolated margin on Hyperliquid is not cosmetic — it directly sets your liquidation probability across every open leg. Getting it wrong is how a well-modeled strategy blows up on a correlated move it was never designed to absorb.

How Each Mode Works Under the Hood

In cross margin mode, Hyperliquid's risk engine draws on your full available account balance when calculating whether a position is solvent. If BTC-PERP runs against you, unrealised PnL from a profitable ETH-PERP leg offsets the loss in real time. The liquidation trigger is evaluated against total account equity, not per-position equity.

In isolated margin mode, each position has its own ring-fenced collateral bucket. The engine sets a maximum loss equal to the margin allocated to that slot. If the position hits zero margin, it is liquidated — full stop — and the rest of your account is never touched.

One subtlety: on Hyperliquid, the cross-margin pool also supports portfolio margining across correlated assets. Long BTC and short ETH in cross mode will often produce a lower combined margin requirement than two isolated positions, because the engine recognises the partial hedge. This is explicit in their margin calculation, not implied.

The Six Scenarios

Running cross margin vs isolated margin through six representative portfolio shapes reveals a clear pattern.

Scenario 1 — Single directional trade, large size. If you are putting on one large bet with no offsetting position, cross margin exposes your entire account to that one outcome. Isolated margin with a defined allocation is strictly better: you know exactly what you are risking.

Scenario 2 — Delta-neutral pair trade (long BTC / short ETH). Cross margin wins here. The engine sees the partial hedge and reduces the combined margin requirement. In isolated mode you are posting full initial margin on both legs and receiving no credit for the correlation. Capital efficiency drops by roughly 20–35% depending on correlation.

Scenario 3 — Funding harvest (large delta-neutral notional, thin margin). This is where cross margin gets dangerous. You are running high notional to collect small funding payments. One leg blowing through its projected range can consume account equity fast, and the other leg's unrealised loss is now also draining the pool. We have seen accounts liquidated during flash crashes in cross mode that would have survived in isolated — the total equity crossed zero before the rate-limit on orders could flatten exposure.

Scenario 4 — Multi-coin directional basket (long SOL, BTC, ARB simultaneously). Cross margin looks attractive but is treacherous: these assets are correlated. A risk-off event hits all three at once. The cross pool is absorbing three simultaneous drawdowns. A single isolated position might survive a 15% move; three correlated longs in cross mode liquidate together. Use isolated for each leg and size each independently.

Scenario 5 — Stat-arb spread trade with tight entry, wide stop. Cross margin is correct. The strategy's edge depends on spread mean-reversion; individual legs move violently during spread widening. Isolated mode will liquidate the losing leg before the spread closes. Cross mode keeps both legs alive and lets the trade work.

Scenario 6 — Experimental position alongside a core book. Isolated. Always. You want the experiment capped at exactly what you allocated — not able to eat into the equity supporting your main book.

The Liquidation Math That Actually Matters

Hyperliquid liquidates a cross-margin account when account equity falls below maintenance margin across all open positions combined. The formula is straightforward:

Liquidation threshold = Σ (|position_notional_i| × maintenance_margin_rate_i)

If account equity ≤ that sum, the engine begins liquidating the largest losing positions first.

In isolated mode, each position has its own threshold:

Liquidation price = entry_price × (1 − initial_margin_fraction + maintenance_margin_fraction)

for a long; the mirror for a short. The rest of your balance is never part of this calculation.

The practical implication: in cross mode, a profitable position raises your liquidation buffer. In isolated mode it does nothing for the other slots. This is the core trade-off — shared buffer is capital-efficient but introduces contagion between unrelated bets.

When the Bot Decides

If you are running an automated Hyperliquid perps bot that opens and closes positions programmatically, the margin mode needs to be a first-class configuration parameter, not an afterthought. At TierZero, our production Hyperliquid bots select margin mode at the strategy level:

• Market-making and funding arb legs run in cross margin so correlated hedges receive portfolio margin treatment and capital is not stranded in isolated buckets.
• Signal-driven directional positions always run isolated, with the allocated margin sized explicitly in the strategy config. This ensures a bad signal cannot consume equity allocated to the market-making book running in parallel.
• Experimental signals during live testing are always isolated at 1–2% of total account equity, enforced by the bot's pre-order validation layer.

Switching margin mode mid-trade on Hyperliquid closes and reopens the position, which means a round-trip fee and potential slippage. Bake the mode into strategy initialisation — do not toggle it reactively.

What the Scenarios Say in Summary

Portfolio shape	Better mode	Key reason
Single large directional bet	Isolated	Defined max loss
Delta-neutral pair trade	Cross	Portfolio margin credit
Funding harvest, thin margin	Isolated	Contagion risk in correlated crash
Multi-coin correlated basket	Isolated per leg	No cross-pool contagion
Stat-arb spread trade	Cross	Prevents premature leg liquidation
Experiment alongside core book	Isolated	Hard cap on experiment loss

The pattern: use cross when your positions are explicitly hedged and the engine's correlation model works in your favour. Use isolated when positions are independent bets or correlated in the same direction — contagion risk in a cross pool outweighs the capital-efficiency gain.

Neither mode is universally superior. The question is whether your open positions provide genuine mutual insurance or whether they all go wrong at the same time.

---

If you are building a Hyperliquid strategy and want the margin logic handled correctly at the infrastructure level rather than patched after a blowup, get in touch — we ship these systems in production.