PERPS ATLASPERPETUAL MARKETS INTELLIGENCE
PERP DEX DIRECTORY
Home / Research / Perp Dex Liquidation Risk
← Research library
TRADING METHODS

Perp DEX liquidation explained: collateral, margin and funding

A perpetual contract has no expiry, but a leveraged position can close when its equity falls below a venue-specific maintenance threshold. This guide follows the collateral, trigger price and holding costs that move that boundary.

By PerpsAtlas Research · Published · Updated · 5 min read

Start with the collateral buffer

Collateral supports the trade; notional is the market exposure. Supplying 1,000 USDC for a 10,000 USDC long creates 10× initial leverage. A 1% adverse move changes unrealized profit and loss by roughly 100 USDC before fees. The trader has price exposure through a contract, not ownership of 10,000 USDC of the underlying asset. Hyperliquid states its opening margin requirement as position value divided by selected leverage; Ostium likewise defines notional from collateral and leverage.

Opening margin determines whether an order may be placed. Maintenance margin is the minimum equity needed to keep a position open. Equity changes with unrealized profit or loss and applicable payments. A displayed liquidation price is therefore a risk boundary, not a guaranteed exit price. Hyperliquid says the displayed figure may change with funding and, under cross margin, profit or loss elsewhere in the account.

Sources: Hyperliquid margining ↗ · Hyperliquid liquidations ↗ · Ostium opening a trade ↗

Which balance is at risk?

With isolated margin, collateral is assigned to a particular position. Adding eligible collateral gives it more room, while removing collateral narrows the buffer. With cross margin, positions share a margin pool, so an unrelated loss can weaken the same account. Hyperliquid also distinguishes account abstraction modes: unified accounts share a balance across eligible books using the same collateral; standard accounts separate DEX balances. Portfolio margin uses an account-wide test that can include eligible spot assets and borrowing, so the isolated example below must not be applied to it.

Portfolio margin is conditional rather than a default promise. Hyperliquid documents eligibility thresholds, asset and borrowing caps, and fallback to non-portfolio behavior when caps are reached. Its account-wide margin ratio is tested near 95%; the technical guide says greater than 0.95, while its FAQ says at or above 95%, so the exact boundary should not be inferred from this article. Eligible spot collateral and borrowing can affect the test; borrowing accrues interest, and subaccounts remain separate. The order of oracle updates can affect whether a perp position or spot borrowing is addressed first. A trader should identify the actual account mode and collateral before interpreting one position's displayed leverage as the whole account's risk.

For portfolio margin, the documented backstop liquidator takes over liquidated positions directly; spot-borrow liquidation has no market-order phase. This differs from the ordinary perp-margin sequence described in the worked example below. Hyperliquid also labels the older DEX abstraction account mode discontinued.

Sources: Hyperliquid margining ↗ · Hyperliquid account abstraction modes ↗ · Hyperliquid portfolio margin ↗

A worked Hyperliquid BTC liquidation boundary

Assume an isolated BTC long opens at 100,000 USDC per BTC. The trader supplies 1,000 USDC and takes 0.1 BTC of perpetual exposure, worth 10,000 USDC at entry. Assume the position remains in Hyperliquid's documented first BTC tier, whose maximum leverage is 40×. Maintenance margin is half the initial margin rate at maximum leverage: 1.25% of current notional in this first-tier example. Ignore trading fees, funding, tier changes and other adjustments for the first calculation.

At mark price P, simplified equity is 1,000 + 0.1 times (P minus 100,000) USDC. Simplified maintenance margin is 0.0125 times 0.1P. Equating them gives P = 91,139.24 USDC, an 8.86% fall from entry. This is an illustrative model boundary, not a live liquidation quote or an executable sale price. Hyperliquid triggers ordinary perp-margin liquidation using its mark price, which can differ from the book price. For ordinary perp margin it first attempts an order-book close; a separate backstop path applies under deeper stress. Portfolio margin follows the direct backstop process described above.

Sources: Hyperliquid margin tiers ↗ · Hyperliquid margining ↗ · Hyperliquid liquidations ↗ · Hyperliquid robust price indices ↗

Ostium uses a different threshold

Ostium's current liquidation guide defines threshold loss as 100% minus (chosen leverage divided by pair maximum leverage times 25%). Its example of a 20-times BTC trade with a 200-times pair maximum gives 97.5% loss of collateral, or approximately a 4.875% adverse price move before opening and oracle fees, rollover, and execution effects. Ostium says accrued rollover fees reduce effective collateral, while opening and oracle fees are deducted at entry; a real trade therefore need not start with the full displayed deposit as effective collateral.

The two venues also describe different liquidation execution and settlement. Ostium says its keeper submits the transaction when the mid-price crosses the threshold, and the remaining collateral is retained as part of settlement. Ordinary Hyperliquid perp margin uses a mark price, initially attempts book liquidation and can leave residual collateral after a successful partial or full book close. The Hyperliquid illustration uses 10-times leverage and the Ostium example uses 20-times leverage: their 8.86% and 4.875% moves are not a like-for-like venue comparison. Execution also depends on bid/ask spread and price impact.

Sources: Ostium liquidation ↗ · Ostium opening a trade ↗ · Ostium fees ↗ · Hyperliquid liquidations ↗ · Ostium markets ↗

Holding costs can move the boundary

Hyperliquid funding is a periodic transfer between long and short holders. A positive rate means longs pay shorts; a negative rate reverses the direction. It settles hourly, using position size, oracle price and the interval's funding rate. If the example long paid a hypothetical cumulative 10 USDC funding debit, replace 1,000 with 990 in the equity equation. Holding the other assumptions fixed, the model boundary rises to 91,240.51 USDC per BTC, a 101.27 USDC shift. Actual funding rates and oracle notional vary, so this is not a forecast.

Ostium's current documentation describes variable rollover on all pairs, including crypto, accruing continuously per block. One side may receive rather than pay, but it is not Hyperliquid-style zero-sum funding between Ostium longs and shorts. Ostium also lists opening and oracle charges and a conditional early-close fee on profitable positions closed within 15 seconds. An older Ostium GitBook describes different mechanics and is marked deprecated; it should not be used to infer current holding cost.

Sources: Hyperliquid funding ↗ · Ostium fees ↗ · Deprecated Ostium GitBook ↗

Five checks before placing the trade

Check the collateral actually assigned after entry charges, the account or position margin mode, the current maintenance rule, the price used to trigger liquidation and the live holding-cost rate. A displayed estimate or advertised schedule is only part of the answer. Executable entry and exit also depend on spread, price impact and completed fills; a stop-loss can execute with slippage during volatility. The account's live position record is needed to know what happened on a particular trade.

Our examples use venue-published rules checked on 7 October 2026 and explicitly stated assumptions. This is the review date, not a claimed effective date for each rule. They do not use PerpsAtlas volume, revenue or token-price observations to infer liquidation risk. A protocol's growth and its token's value are separate questions, neither answered by a worked trader-risk example.

Sources: Hyperliquid liquidations ↗ · Ostium liquidation ↗ · Ostium fees ↗

What we monitor next

  • Recheck live venue documentation and pair-specific margin tiers before reusing these illustrative boundaries.
  • Use account records and executed fills, rather than an advertised schedule, to reconstruct a real position's costs and outcome.
  • Keep current Ostium rollover rules separate from deprecated GitBook funding descriptions.

Frequently asked questions

Does 10× leverage mean liquidation after exactly a 10% price move?

No. Maintenance margin, entry and holding costs, mark or mid-price rules and account mode change the boundary.

Can funding change my liquidation price?

Yes. A funding debit or credit changes account equity. The worked 10 USDC debit is hypothetical, not a current rate.

Do all perp DEXs charge the same holding cost?

No. Hyperliquid documents periodic peer-to-peer funding; Ostium's current documentation describes a variable rollover mechanism.

Does a portfolio-margin liquidation always start with a market order?

No. Hyperliquid documents direct backstop takeover for portfolio margin, including spot-borrow liquidation without a market-order phase. The order-book-first description applies to ordinary perp margin.

Compare the evidence

Open the comparison desk →

Related research

Source register

  1. Hyperliquid margining · checked 2026-09-30
  2. Hyperliquid liquidations · checked 2026-09-30
  3. Hyperliquid margin tiers · checked 2026-09-30
  4. Hyperliquid account abstraction modes · checked 2026-09-30
  5. Hyperliquid portfolio margin · checked 2026-09-30
  6. Hyperliquid funding · checked 2026-09-30
  7. Hyperliquid robust price indices · checked 2026-09-30
  8. Ostium opening a trade · checked 2026-09-30
  9. Ostium liquidation · checked 2026-09-30
  10. Ostium fees · checked 2026-09-30
  11. Deprecated Ostium GitBook · checked 2026-09-30
  12. Ostium markets · checked 2026-10-07
Research revisions
  • 2026-09-30 — First source-reviewed edition; distinguished venue liquidation rules and hypothetical funding effect.
  • 2026-10-07 — Expanded portfolio-margin and spot-borrow mechanics, distinguished ordinary perp liquidation, and clarified the limits of cross-venue examples.

AI-assisted research checked against cited sources. Facts, assumptions and interpretation are distinguished; this is not a financial audit or a recommendation tailored to you. Editorial standards.

Compare protocols →