Hyperliquid Partial Liquidation Mechanics: The Math Behind Why You Lose 30% of Capital but Keep 70% of Risk

A trader opens a 5x leveraged long position on Hyperliquid’s perpetuals market with $10,000 in collateral, controlling $50,000 notional exposure. The asset rises 4%, creating a $2,000 paper gain. Then the market reverses 6%, wiping out the gain and eroding capital. The liquidation engine activates, triggering a partial liquidation that closes 30% of the position—$15,000 notional—and removes $3,000 in collateral. The trader is relieved: they were not liquidated entirely. But the mathematics reveal a harder reality. Before the liquidation, the position risked $10,000 to control $50,000. After, it risked $7,000 to control $35,000. The capital-to-risk ratio deteriorated from 1:5 to 1:5. The position was reduced, but its fundamental instability remained. Understanding this mechanic is essential for anyone using leverage on Hyperliquid or any decentralized exchange.

Partial liquidation is a mechanism designed to preserve capital—both the trader’s and the protocol’s—by closing a portion of an underwater position rather than the entire thing. On the surface, it appears protective. In practice, it creates a mathematical trap: survivors often face worse odds than before liquidation, unaware that their capital has not been salvaged so much as repositioned into a more precarious state. The liquidation calculator and liquidation price estimator tools available on the platform can help traders model these scenarios, but they require understanding the underlying mechanics to use them correctly. This article examines how partial liquidation actually works on Hyperliquid, why the survivors often carry amplified risk, and how to model the true cost of being partially liquidated.

The partial liquidation mechanism on Hyperliquid

Hyperliquid operates its own L1 blockchain with an on-chain order book, which means liquidation decisions and position closures are processed directly on-chain rather than through an intermediary settlement layer. When a position falls below its liquidation price, the protocol does not immediately close the entire exposure. Instead, it calculates a partial closure: typically removing enough notional value to restore the position to a healthier collateral ratio, leaving the remainder open.

The exact amount closed depends on the position’s margin mode. In cross margin, all available collateral across the account funds all positions collectively. In isolated margin, each position has its own collateral pool. Strict isolated margin locks collateral even more tightly. The liquidation engine evaluates the current collateral level against the maintenance margin requirement—typically more lenient than the initial margin requirement that triggered the opening of the position. If the account (or isolated position) falls below maintenance, a partial close occurs to push it back above that threshold.

Consider a concrete example using Hyperliquid’s infrastructure. A trader with $20,000 collateral opens a 10x long position in ETH, controlling $200,000 notional. The maintenance margin requirement is 5%, meaning the position needs at least $10,000 backing it ($200,000 × 0.05). If ETH drops 4.75%, the notional falls to $190,500, and the loss is $9,500, leaving collateral at $10,500. The account is still above the 5% maintenance threshold. But if ETH drops another 0.5% (5.25% total), the position is worth $189,500, the loss is $10,500, and collateral falls to $9,500—below the $10,000 maintenance requirement. The partial liquidation mechanism then activates, closing enough of the position to restore the ratio.

The protocol calculates the closure as follows: if $9,500 collateral remains and the maintenance requirement is 5%, the maximum sustainable notional is $190,000 ($9,500 ÷ 0.05). The current notional is $189,500, so the position is already close to the threshold but not yet breached; however, the liquidation price has been touched. The liquidation engine closes notional equal to a haircut—typically calculated to leave a margin of safety. If it closes 20% of the position ($37,900 notional), the remaining exposure is $151,600. With $9,500 collateral, the collateral ratio is now 6.26% ($9,500 ÷ $151,600), above the 5% maintenance level. The partial close succeeds, and the trader retains the remaining 80% of the position.

Why the capital-to-risk ratio worsens, not improves

The counterintuitive result of partial liquidation is that the surviving position often becomes riskier per dollar of remaining collateral, despite having lower absolute leverage. This occurs because partial liquidation closes the position at the worst possible price—precisely when the market has already moved against the trader and when liquidation conditions are activated. The closed portion generates a realized loss. That loss is deducted from collateral. The remaining position, however, keeps its original leverage multiple relative to the now-smaller collateral base.

Return to the example: before liquidation, the trader had $20,000 collateral supporting $200,000 notional (10x leverage). After partial liquidation closes 20%, collateral has fallen to $9,500 (the $10,500 loss being $9,500), and notional is $151,600. The remaining leverage is now $151,600 ÷ $9,500 = 15.96x. The leverage has risen, not fallen. The trader started at 10x and is now at approximately 16x, despite having less absolute exposure. This is the crux of the trap: partial liquidation simultaneously reduces the trader’s buffer and increases the leverage multiple of what remains.

The reason is structural. The position did not voluntarily de-lever. It did not sell at a good price or reduce size methodically. Instead, the protocol forcibly closed the weakest tranche—the one that had already suffered the steepest drawdown—at the worst time. The surviving position is smaller in notional but also smaller in collateral coverage. If the market continues to move against the trader, the remaining position will approach its new liquidation price faster than the old position would have. The distance to liquidation has shrunk faster than the position size.

To quantify this, traders can use a liquidation calculator to model the exact mechanics for their chosen asset and leverage. The tool shows that as collateral shrinks, even a small additional adverse price move triggers another liquidation event. This creates a liquidation cascade risk: the first partial close leaves the position vulnerable to a second partial close, which leaves it vulnerable to a third, until the entire position is extinguished. The survivor is not safer. It is one step closer to total liquidation at a worse-than-before capital ratio.

The liquidation pricing penalty

When Hyperliquid’s on-chain order book processes a partial liquidation, the closed portion is executed against the current market price or worse. Unlike a limit order placed by a human trader, a liquidation close is a market order: it hits the bid (for long closes) or ask (for short closes) immediately. If the market is moving quickly or if there is insufficient liquidity at the current level, the close may execute several basis points or even percentage points worse than the last recorded price.

The protocol’s liquidation price estimator can provide a theoretical liquidation level, but the actual fill price when liquidation occurs may differ. This slippage is a second hidden cost of partial liquidation. A position that hit its theoretical liquidation price might close at a 0.5% worse price than anticipated, adding hundreds or thousands of dollars in loss to accounts with large notional exposures. That loss comes directly from the surviving collateral, making the remaining position even more fragile.

The role of the HLP (Hyperliquid Liquidation Pool) further complicates pricing. The HLP functions as a backstop for liquidated positions, stepping in when the standard order book does not have enough liquidity. If the HLP needs to absorb a liquidation, it may require a wider spread to account for the inventory risk. This is not a flaw in Hyperliquid’s system—it is a feature that prevents a liquidity crisis. However, it means that traders liquidated during high-volatility periods or when the HLP is already stressed face worse execution than those liquidated during calm markets.

Margin modes amplify the trap differently

Hyperliquid offers three margin modes: cross, isolated, and strict isolated. Each changes how partial liquidation affects capital and risk allocation.

In cross margin mode, all available collateral backs all open positions collectively. A partial liquidation on one position releases collateral that can support other positions, but it also means a large loss on one trade can reduce the buffer for unrelated trades. If a trader has ten small positions and one heavily leveraged position that gets partially liquidated, the loss is distributed across the collateral pool. The trader may not notice the damage until trying to open a new position and finding less capital available.

Isolated margin mode assigns collateral to individual positions. A partial liquidation on an isolated position only affects that position’s buffer. This reduces contagion to other trades, but it also means the isolated position faces the full brunt of the capital erosion without any offset from unrelated activity. A 10x isolated position that gets partially liquidated drops to 16x or higher, as calculated earlier, with no collateral from other positions to cushion further losses.

Strict isolated margin is the most restrictive. It locks collateral into a position until the position is closed. Partial liquidation of a strict isolated position not only reduces the capital buffer but also prevents the trader from accessing that collateral for any other opportunity. The locked-in capital becomes a sunk psychological cost as well as a mathematical one.

The choice of margin mode therefore determines not just how much capital is at risk, but how that capital is sequestered during and after a partial liquidation. Traders who default to cross margin without understanding its contagion effects often discover—via partial liquidation—that a single loss has undermined their entire portfolio leverage. Traders who use isolated margin on high-conviction trades often discover that isolation works both ways: it protects other positions but leaves them personally bearing the full weight of de-leveraging.

Real-world partial liquidation scenarios on Hyperliquid

Scenario 1: The 5x Bitcoin long. A trader deposits $5,000 USDC via Arbitrum (the minimum is 5 USDC, so this is well above it) and opens a 5x long on BTC at $42,000. Position size: $25,000 notional. Maintenance margin: 5%, so the minimum collateral needed is $1,250. Current buffer: $3,750 above maintenance. BTC drops to $39,900 (a 5% decline). Notional loss: $1,250. Collateral is now $3,750. The position is exactly at maintenance. A partial close is triggered, removing 20% of notional ($5,000). New notional: $20,000. New leverage: $20,000 ÷ $3,750 = 5.33x. The trader is still at roughly 5x, but with $1,250 less cushion. If BTC falls another 3%, another partial close will likely activate.

Scenario 2: The 10x ETH short with isolated margin. A trader uses isolated margin to short ETH at $2,250 with 10x leverage. Collateral: $2,000. Notional: $20,000. ETH rises to $2,363 (a 5% move). Notional loss: $1,000. Collateral: $1,000. The position is at 5% maintenance, so a partial close occurs, removing 25% of the short ($5,000 notional). New notional: $15,000. New leverage: $15,000 ÷ $1,000 = 15x. The trader is now carrying 15x leverage on $1,000, with only a 1% price move ($225) away from total liquidation. If ETH rises another 1%, the entire position is liquidated at market price, likely with additional slippage.

Scenario 3: The cross-margin cascade. A trader holds five isolated positions totaling $50,000 collateral. One position—a 20x alt-coin long—gets partially liquidated due to a 4% drop, losing $4,000. The collateral falls to $46,000. The trader does not notice and opens another position. But when the market reverses sharply, multiple positions near their liquidation prices begin to cascade. Each partial liquidation removes collateral, making the next liquidation more likely. Within minutes, three positions are partially liquidated, and the trader has lost $12,000 in capital, leaving $38,000. The surviving positions are now at higher leverage multiples than when opened, and another sharp move will finish them off.

Each scenario illustrates the same mechanic: partial liquidation stops the bleeding momentarily but leaves the wound worse than before. The liquidation calculator on hyperliquid allows traders to model these exact scenarios beforehand, inserting their leverage, collateral, and asset price to see where liquidation would trigger and what leverage would remain. Using this tool is not optional for traders intending to use 5x or higher leverage. It is the difference between understanding the risk and discovering it during a market move.

Fee structures and the cost of survival

Hyperliquid’s base taker fee for perpetuals is 0.045%, and the maker fee is 0.015%. These are charged on notional traded, not on collateral. During a partial liquidation, the protocol executes a taker trade to close the position. A trader whose $20,000 position is 20% closed—$4,000 notional—pays a 0.045% fee on that closure, or $1.80. For large positions, this is trivial. For accounts operating near their margin limits, every basis point matters.

Moreover, traders can reduce fees through volume tiers and HYPE token staking. A trader who has not staked HYPE and is at a lower volume tier pays the full 0.045% taker rate on liquidation closes. One who has staked and reached a higher tier pays less. However, the fee structure creates a perverse incentive: traders who lose money and get partially liquidated are often those with lower volumes and less HYPE staked, meaning they pay the highest fees on their worst trades. Those who are winning and trading frequently get fee discounts, creating a wider spread between winners and losers.

The referral structure offers another layer of nuance. Traders who have referred others can earn discounts on fees. Those referred by a strong referral account may also receive discounts. However, these are benefits accrued before liquidation. Once liquidation begins, the fee is assessed on the market order executed by the protocol, regardless of referral status.

De-risking after a partial liquidation

After a partial liquidation, the survivor faces a critical choice: close the remaining position or hold and hope for recovery. Closing immediately locks in the total loss and removes the position from leverage. Holding retains exposure but also retains the amplified leverage of the now-smaller collateral base.

The mathematical argument for closing is straightforward: after a partial liquidation, the position has moved against the trader hard enough to trigger forced de-leveraging by the protocol. The surviving position is now at a higher leverage multiple than was prudent enough to survive that move before. Closing removes it from the risk of cascade liquidation. The cost is admitting the full loss and accepting that the remaining capital is now smaller.

The psychological argument for holding is equally familiar: the trader believes the move is temporary and that reopening at the original price will profit from the recovery. However, this ignores a crucial fact. The remaining position at 15x or 16x leverage is far less forgiving than the original 10x position was. A market recovery of 1% that would have recovered half the loss on the original position may only save the revised position from another partial close. The math has fundamentally changed.

A practical framework: after any partial liquidation, the trader should immediately recalculate the new liquidation price using a liquidation price estimator and assess whether holding is worth the risk given the reduced buffer. If the new liquidation price is only 2-3% away and the market is volatile, holding is a bet on a very narrow recovery. If it is 5-10% away, the position has more breathing room, but that breathing room is still less than the original position had. The decision should be deliberate, not emotional.

Systemic implications and protocol incentives

Hyperliquid’s partial liquidation design reflects a broader principle: liquidation is meant to protect the protocol and other users from cascading defaults, not to minimize individual trader losses. This is appropriate. However, it creates an incentive structure where the protocol benefits from volatility (more liquidations = more closed positions = more protocol revenue and backstop fund activity) and traders suffer. This is not unique to Hyperliquid; the Bank for International Settlements has documented similar dynamics in centralized derivatives markets, noting that liquidation cascades and leverage amplification are systemic risks that regulators monitor globally.

Regulatory bodies including the Financial Conduct Authority have issued consumer warnings about leverage products, specifically flagging partial liquidations and margin calls as hidden costs that ordinary retail traders often misunderstand. The broader point is that Hyperliquid’s mechanics are not aberrations; they are standard in decentralized derivatives. Understanding them is critical because there is no backstop protection if a trader loses more than their collateral (though on Hyperliquid, the on-chain model makes this extremely unlikely, as liquidation is automatic).