A trader opens a 50x leveraged long position on ETH, depositing $1,000 in collateral to control $50,000 notional exposure. The market moves 2% against their position. On a centralized exchange with full liquidation, that account is closed entirely and the margin balance is seized. On Hyperliquid, the same 2% move may trigger a partial liquidation that closes only enough of the position to restore safe margin levels—preserving 80% or 90% of the original position so the trader can keep their exposure and potential upside. Understanding why partial liquidation exists, how it works mechanically, and where the risks still lie is essential for anyone trading leveraged derivatives at scale.

The distinction between partial and full liquidation is not merely a feature difference. It reflects a fundamental choice about who bears the cost when leverage collides with volatility. Full liquidation transfers all remaining value to the exchange or liquidators; partial liquidation leaves the trader holding a reduced but still profitable or recoverable position. This matters because leverage amplifies both gains and losses. A position that survives liquidation with partial loss intact can be re-sized, hedged, or closed at a chosen price rather than a forced one. Hyperliquid’s implementation of partial liquidation on its own blockchain creates concrete operational scenarios that differ sharply from exchange designs where closure means total loss.

How partial liquidation differs from total loss models

Traditional centralized exchanges and some decentralized platforms use a binary liquidation model: margin ratio falls below a threshold, the position is closed at market or a liquidation price, and whatever remains after fees and loan repayment is returned to the user—often nothing. This design simplifies risk accounting for the platform. The exchange or liquidators know exactly when the event occurs, can execute a clean closure, and reduce their own exposure in a single transaction. The user loses the position, the collateral, or both, depending on the size of the move and the severity of slippage during forced execution.

Partial liquidation operates differently. Instead of closing the entire position when margin falls below the liquidation threshold, the system closes the minimum amount necessary to bring margin levels back above the safety line. If a trader has a position worth $50,000 notional and the margin level falls to 5%, a partial liquidation might close $5,000 to $10,000 of the exposure, leaving $40,000 to $45,000 still open. The trader can then observe the market, add collateral if they choose, or wait for price to recover. The position remains on the books and can be managed rather than erased.

This approach imposes more complexity on the exchange’s infrastructure. Rather than a simple «liquidate everything» algorithm, the system must calculate the exact amount to close, execute that portion at favorable prices if possible, and update the margin accounting to reflect the smaller remaining position. Hyperliquid’s architecture, which operates its own L1 blockchain and on-chain order book, is designed to handle this granularity. The platform does not rely on a centralized liquidation engine; it processes liquidations directly on-chain, where order matching and price discovery happen against real liquidity rather than a formula or backstop price.

The consequence is that traders face liquidation risk rather than liquidation certainty. A 5% adverse move no longer automatically equals total account erasure. Instead, it equals a partial position closure at a known or discoverable price. A 10% move may still trigger additional partial liquidations, progressively shrinking the position. Traders can respond between liquidation events by depositing more collateral, closing the remainder voluntarily, or rebalancing across their portfolio.

Margin modes and partial liquidation mechanics on Hyperliquid

Hyperliquid offers three margin modes: cross, isolated, and strict isolated. Each mode changes how collateral is shared and how partial liquidation is calculated. In cross margin mode, all collateral in the account backs all positions simultaneously. A loss on one position can be offset by gains on another, and the liquidation threshold is calculated against the total account margin ratio. If the ratio falls below the minimum, the system closes the smallest positions first—those with the highest ratio of notional value to margin contributed. This creates a form of triage where the system tries to preserve larger, more strategic positions while sacrificing smaller ones.

Isolated margin mode segregates collateral. A trader allocates specific USDC to a specific position; that position uses only its own collateral for margin calculation and liquidation. If a BTC position has $2,000 allocated and its margin ratio falls below the liquidation threshold, only that position can be liquidated—the trader’s ETH, SOL, or other positions remain untouched. This design reduces contagion but increases the number of liquidation decisions the system must monitor. A trader managing five isolated positions is essentially running five mini-accounts, each with its own margin level and liquidation trigger.

Strict isolated margin is the most conservative. Not only is collateral segregated, but the maximum leverage available is reduced. A trader cannot use as much leverage on a single isolated position as they could in cross mode on the same collateral, making liquidation less likely per unit of notional exposure. This mode appeals to risk-averse traders or those building positions they intend to hold through volatile periods without close monitoring.

Partial liquidation works within each mode. In cross margin, the system may close 20% of the aggregate notional exposure. In isolated margin, it closes a portion of the specific position that triggered the liquidation. The hyperliquid liquidation calculator helps traders model these scenarios before trading: enter a position size, leverage, collateral, and assumed price move, and the tool shows the liquidation price, the amount that would be closed, and the margin level after the partial liquidation. This transparency reduces the surprise factor and allows traders to size positions defensively.

Real-world scenario: 2% move, partial liquidation, trader survives

Consider a concrete example. A trader deposits $5,000 USDC and opens a long ETH position using 25x leverage in cross margin. The notional exposure is $125,000. ETH is trading at $2,500. The position is 50 ETH. The minimum margin level for liquidation is 5%. Initial margin is $5,000 / $125,000 = 4%, which is already at the edge—this position is risky from the start.

ETH drops 2%, to $2,450. The position is now worth $122,500, a $2,500 loss. Margin level is now ($5,000 – $2,500) / $122,500 = 2%. This breaches the liquidation threshold. Instead of closing the entire 50 ETH position and leaving the trader with nothing, Hyperliquid’s partial liquidation closes enough to restore margin to approximately 5% to 6%. To get to 5% margin, the system needs to close positions until ($5,000 – $2,500) / notional = 0.05. Solving for notional: $2,500 / 0.05 = $50,000. So the system closes $125,000 – $50,000 = $75,000 of notional exposure, which is 30 ETH. The trader is left with 20 ETH still long.

The trader’s account now shows: 20 ETH position, $2,500 margin remaining, margin level 5%. If ETH stays at $2,450, the trader can now do nothing and hold the position indefinitely. If they add $1,000 more collateral, the margin level jumps to 7%, and they have room to weather further drops. If they close the remaining 20 ETH at $2,450 (the price where the partial liquidation occurred), they receive $49,000, bringing their total remaining balance to $49,000 + $2,500 = $51,500. They lost $5,000 on a 2% move with 25x leverage—a 50% loss on capital, which is severe but not total ruin. On a traditional full-liquidation exchange, the same 2% move would have left them with zero.

If ETH instead recovers to $2,550 after the partial liquidation, the trader’s 20 ETH position is now worth $51,000, the margin level is ($2,500 + $1,000) / $51,000 ≈ 7%, and they still have optionality. They can close for a profit, hold for a larger recovery, or add collateral to re-enter the position size they had before. The key difference is that they retain agency. The market did not erase their account unilaterally.

The role of leverage, liquidation prices, and risk management

Leverage is the primary driver of liquidation risk. High leverage means small adverse moves trigger liquidation events. Low leverage means larger moves are required. Hyperliquid offers variable maximum leverage by asset—Bitcoin and Ethereum may allow up to 100x, while less liquid or more volatile altcoins are capped at lower multiples. This graduated structure reflects the practical reality that liquidating a 100x Ethereum position is more likely to complete at close to the mark price, while liquidating a 100x illiquid altcoin position might incur severe slippage and leave the liquidators holding losses.

The liquidation price is the spot price at which the margin level would breach the liquidation threshold if no other variables changed. For a $5,000 balance, $125,000 notional long ETH position at $2,500 with 5% minimum margin, the liquidation price is calculated backward: at what price does margin ratio equal 5%? ($5,000 – loss) / (notional at that price) = 0.05. If the liquidation price is $2,375, then at $2,375 the position has lost $5,000 (the full collateral), and liquidation occurs. The distance from entry ($2,500) to liquidation ($2,375) is $125, or 5%. Using Hyperliquid’s liquidation price estimator, traders can confirm these calculations before trading and adjust position size or leverage to create more buffer.

Risk management therefore has three levers. First, reduce leverage: 10x is safer than 25x. Second, increase collateral relative to notional exposure: a $5,000 balance with a $50,000 position (10x) has more margin buffer than a $5,000 balance with a $125,000 position (25x). Third, use isolated margin for positions where losses should not threaten other holdings, and use strict isolated mode if the goal is maximum safety at the cost of lower leverage. Many professional traders combine techniques: they may use cross margin on a small core position they are confident in, isolated margin on speculative trades, and strict isolated for any position they expect to hold through the night without monitoring.

Liquidation backstops and the HLP mechanism

Partial liquidation still depends on being able to execute a trade. If the market is moving faster than liquidity can absorb, closing a portion of a position might also incur slippage, and the closed portion might not fully offset the margin shortfall. Hyperliquid’s HLP (Hyperliquid Liquidity Provider) backstop and auto-deleveraging (ADL) mechanism address this scenario. If the on-chain order book cannot fill a liquidation order at reasonable slippage, the HLP pool provides the counterside—the liquidators or protocol cover the gap and take the directional risk. This ensures that partial liquidations execute cleanly and do not cascade into systemic insolvency.

The HLP pool is funded by traders who stake USDC into the liquidity provision system in exchange for a share of liquidation profits and trading fees. When a large liquidation occurs and the regular order book cannot absorb it, the HLP becomes the buyer (for short liquidations) or seller (for long liquidations). In return, if the position recovers after liquidation, the HLP pockets the difference. If it deteriorates further, the HLP absorbs the loss. This structure creates a counterincentive: liquidators are motivated to liquidate early and often because the HLP backstop removes the risk of getting stuck in an unwind. For traders, it means partial liquidation is more likely to execute at a known price rather than at a panic-slippage price.

Auto-deleveraging (ADL) is the mechanism of last resort. If the HLP pool is depleted or insufficient to cover a liquidation, rather than force all traders in the account to close involuntarily, the protocol progressively reduces the positions of traders on the opposite side of the market who are most profitable on that trade. A massively profitable long position is partially closed to release collateral that offsets a being-liquidated long position. This sounds adversarial—and it is, in the sense that profitable traders take a loss to save the system—but it is preferable to a cascading collapse where every account gets wiped.

Fee structures, volume tiers, and the cost of liquidation recovery

Trading fees on Hyperliquid start at 0.045% taker and 0.015% maker for perpetuals. For a trader who is liquidated and then re-enters a position, those fees compound the recovery cost. If a $125,000 position is partially closed via liquidation and the trader later re-enters a $75,000 position, they pay 0.045% = $33.75 in taker fees on re-entry, on top of the losses from the liquidation event itself. For smaller traders or those liquidated frequently, these fees represent a second tax on leverage.

Volume-based tier reductions are available: accumulating $100,000 in 30-day volume may reduce fees to 0.040% taker. Staking HYPE tokens provides another fee discount. A trader with HYPE in their account may reduce maker fees to as low as 0.005% or lower, depending on the staking amount. This creates an incentive to hold the protocol’s native token, but it also means that a trader’s effective costs depend on their commitment to the platform, not just their single trade size.

For liquidation avoidance, the fee structure reinforces the case for conservative leverage. A trader paying 0.045% per trade can afford fewer losing liquidations before the cumulative fees exceed the original capital. A trader paying 0.015% maker fee by staking HYPE can hold smaller positions longer and re-enter more cheaply. The platform’s design creates a subtle hierarchy where engaged, well-capitalized traders face lower unit costs and thus have more room for error.

Regulatory context and smart-contract risks

Hyperliquid operates as a decentralized exchange on its own L1 blockchain, but it is not exempt from regulatory scrutiny. The CFTC has established a regulatory framework for derivatives and perpetual futures trading that treats leveraged cryptocurrency products as futures contracts subject to U.S. regulation. The platform restricts access from the United States, Ontario, and sanctioned regions to comply with these rules. However, regulatory changes—particularly around leverage limits, segregation of customer funds, and reporting requirements—could alter the product offering.

The Financial Conduct Authority has issued warnings about retail crypto derivatives trading, noting that leveraged cryptocurrency products carry extreme risk and are suitable only for a narrow investor base. The FCA has restricted retail access to leveraged cryptocurrencies in the UK and is monitoring decentralized exchanges for compliance. For European traders, the European Securities and Markets Authority has established cryptocurrency derivatives regulations that set leverage caps and margin requirements designed to limit retail losses.

Smart-contract risk is the second vector. Hyperliquid’s on-chain order book, liquidation calculations, and HLP mechanism are all implemented as smart contracts on the platform’s L1. If a bug exists in the liquidation logic, a trader might be liquidated at the wrong price, their partial liquidation might not execute as intended, or the HLP mechanism might not trigger when it should. Blockchain security standards, as outlined by the Ethereum Foundation, recommend code audits, formal verification, and gradual rollout of new features. Hyperliquid has published audit reports, but the attack surface of a novel L1 blockchain and custom derivatives engine remains non-zero.

Bridge security is also relevant. Traders deposit USDC via Arbitrum, Ethereum, and other chains using a cross-chain bridge. If the bridge has a vulnerability or is compromised, deposits might be lost or stolen. Hyperliquid’s engineering team has hardened the bridge against common attacks, but the theoretical risk exists. For this reason, traders should begin with small test deposits and verify that withdrawals work before moving larger amounts.

When partial liquidation is not enough: cascading liquidations and market dislocations

Partial liquidation is a risk reduction mechanism, not a risk elimination mechanism. In scenarios where the market gaps sharply—a flash crash, a critical news event, or a sudden loss of liquidity—partial liquidation may not protect a trader fully. If ETH drops 5% in one block, and the liquidation system can only close 2% of a position before the price has moved further, the trader experiences a cascade: first partial liquidation at $2,375, second partial liquidation at $2,362, third partial liquidation at $2,349. By the time the dust settles, more of the position is closed than a single liquidation would have required, and the collateral damage is greater.

This scenario is rare but not theoretical. During the 2022 crypto market crash, some perpetual futures exchanges experienced cascading liquidations where entire portfolios evaporated in minutes. Hyperliquid’s HLP backstop and on-chain execution reduce the likelihood of this scenario by ensuring that each partial liquidation executes at a discoverable price rather than at panic slippage. But extreme volatility or a depeg of USDC (the collateral asset) could still overwhelm the system.

The most defensive position is to maintain a large margin buffer and low leverage. A trader with 5x leverage instead of 25x has five times more headroom before liquidation. That trader can afford losses without partial closure. The math is simple: if leverage is lower, adverse moves must be larger to trigger liquidation. The trade-off is lower capital efficiency; a $5,000 balance controls only $25,000 notional rather than $125,000 notional. But for traders who cannot afford to lose their collateral in a liquidation cascade, this sacrifice is worthwhile.