(Black Mirror) The first episode of the last season (Joan is Awful) story: The female lead casually clicked 'agree', and as a result, the platform turned her daily 'legal' activities into a global reality show based on the user agreement that no one looked at closely.
In fact, every industry has its own user agreements. The same goes for Perp; the liquidation rules are the 'user agreement' of this track.
It may not be sexy or eye-catching, but it is very, very important. The same token has different depths on different trading platforms, different K-line trends, and different liquidation mechanisms, leading to completely different results for positions.
Today's examples of two perp DEXs are excellent teaching materials: at the same time, while Binance's pre-market contract quotes did not show a similar magnitude of fluctuation, the XPL price on Hyperliquid was pushed up nearly +200% in about 5 minutes; and the ETH price on Lighter spiked to 5100 dollars.
In extreme market conditions, some rejoice while others lament.
In just one hour, several whales on Hyperliquid cleared counterparties' short orders by driving up prices, collectively profiting nearly $38 million to $46.1 million; among them, the widely watched address 0xb9c...6801e had been lurking with long positions since August 24 and made about $16 million profit just one minute after the 'sweep'; HLP netted about $47,000 in this incident. But the shorts were not so lucky: 0x64a4's XPL shorts faced successive liquidations within minutes, losing about $2 million; 0xC2Cb's shorts were fully liquidated, losing about $4.59 million.
Rather than saying this is a 'manipulation by bad actors', it is more of a result of the liquidation system and market structure working together, providing a new lesson to all crypto players: always pay attention to depth and liquidation mechanisms.
How does liquidation occur?
On any perpetual contract platform, the first thing to understand is: Initial Margin (IM) and Maintenance Margin (MM): IM determines the maximum leverage available; MM is the liquidation line, when the account equity (collateral plus unrealized profits and losses) falls below it, the system will take over the position and enter the liquidation process.
Next, let's look at the price. The price that determines liquidation is not the last transaction price, but the marked price (Mark). It is often determined by external indexes, oracle predictions, and the platform's own order book, and is processed for smoothing and anti-manipulation; while the index price is closer to a pure external weighted spot reference; the latest price (Last) is the most recent transaction on this platform, which is easily influenced by instantaneous sweeping orders.
So when 'account equity < maintenance margin', liquidation is triggered. However, the specifics of liquidation still depend on the platform's own execution mechanism.
Hyperliquid: Let the market eat the liquidation orders.
First, let's look at Hyperliquid. Hyperliquid's liquidation mechanism is: when account equity falls below the maintenance margin (MM), the system prioritizes directly placing liquidation orders into the order book, eating away at the risk in a market-oriented manner.
If it is a large position (e.g., >100,000 USDC), it usually clears 20% first, then cools for about 30 seconds, and then when it triggers, it might handle the entire position. If it cannot be cleared and further deteriorates to a deeper threshold (e.g., equity below 2/3 × MM), it will trigger backup liquidation, taken over by the community-based Liquidator Vault (HLP component). To ensure that 'the safety net can continue', in many cases, this step does not refund the maintenance margin.
The entire process does not incur 'liquidation fees', but in situations where external anchors are weak and depth is poor, the liquidation action itself will further push the price in the same direction for a while, causing a short-term surge.
At the same time, Hyperliquid's marked price is determined by external CEX quotes and its own order book, and when volatility is driven by internal transactions, it can also accelerate the triggering pace. Additionally, isolated and cross margins behave differently during backup stages: when cross margin (Cross) is taken over, it may transfer the entire position and margin together; isolated margin (Isolated) only affects that specific position and its isolated margin.
So let's rewind to this morning to briefly review the Hyperliquid XPL incident. Starting from 5:35, Hyperliquid's XPL buy orders rapidly raised the transaction range by 'sweeping the order book', with the marked price dominated by the platform's matching, jumping far beyond the normal range.
For a side that is heavily crowded with shorts, this spike led to a sudden drop in the ratio of account equity/maintenance margin: when equity falls below MM, the system takes over the position according to established procedures. Next, the system will buy to close the shorts in the order book (larger positions may be partially closed first and enter a cooling period), and this set of short-closing buy orders will continue to push prices and marked prices higher, triggering more shorts to fall below MM. Within seconds to tens of seconds, the positive feedback of 'sweep → trigger liquidation → liquidation sweeps again' is formed—this is also the mechanistic explanation for the price approaching +200% in just a few minutes.
If it still cannot be cleared on the order book, and the equity of the liquidated account further drops to a deeper threshold (like 2/3 × MM), the backup liquidation mechanism will take over, ensuring that the buy orders used to cover the losses continue to come in, ensuring that the risk is 'eaten' within the system.
Once the order book regains depth and the liquidation queue is digested, addresses that actively go long start to realize profits, and the price drops rapidly from the high like a roller coaster—this is the entire process of 'Hyperliquid's XPL rising nearly +200% in a few minutes → successive short liquidations → rapid decline.'
This is actually the inevitable result of the coupling of pre-market liquidity depth + position crowding + liquidation mechanism.
Does Lighter really liquidate earlier?
Now, let's take a look at Lighter. This morning, the ETH price on Lighter spiked to 5100 dollars, which also attracted quite a bit of attention. As a perp DEX that is currently second in trading volume only to Hyperliquid, Lighter's price spikes have often sparked discussions.
Lighter has three thresholds: IM (Initial Margin) > MM (Maintenance Margin) > CO (Close-out Margin). Falling below IM enters 'pre-liquidation', where you can only reduce, not increase; falling below MM enters 'partial liquidation', where the system issues IOC limit orders at the pre-calculated 'zero price' to reduce your position. The design of zero price ensures that even if the transaction is made at zero price, the account health will not continue to deteriorate; if executed at a better price, the system will take a liquidation fee of no more than 1% into the LLP (insurance fund). Further down, equity below CO triggers 'complete liquidation', clearing all positions and transferring remaining collateral to the LLP; if the LLP funds are insufficient to cover, ADL (Auto Deleveraging) will be activated, partially offsetting high-leverage/large unrealized profit positions on the counterpart side at their respective zero prices, aiming to achieve deleveraging at the system level without harming innocents. Overall, Lighter sacrifices some 'liquidation speed' in exchange for controllable account health and book impact.
So does this mean that Lighter's liquidation price is much higher than other DEXs and will liquidate earlier?
A more accurate answer is: yes, but not entirely.
In simpler terms, Lighter's 'earlier liquidation' is a phased 'deleveraging to extinguish fires': it issues IOC limit orders to deleverage at 'zero price', aiming not to worsen account health. Often, it reduces positions to a safe level, which does not mean you will be 'liquidated'; only worsening to CO will trigger a complete liquidation.
Therefore, it cannot simply be said that 'Lighter is easier to liquidate', but rather that it liquidates and deleverages more gently, spreading the risk to reduce the price impact of a single sweep; the cost is that if the transaction is made at a price better than 'zero price', a liquidation fee of ≤1% will be charged into the insurance fund (LLP).
Interestingly, early on, the weight of the liquidation amount in Lighter's points was relatively high. According to analysis by Lighter community member 0xTria, at that time, the initial liquidation of accounts was '1 dollar of liquidation ≈ 1 point', and the community valued points at 15-30 dollars, which directly led many users to exploit sub-accounts and new accounts to 'liquidate for points' for arbitrage. However, in a later version, this weight was significantly reduced.
How to avoid being manipulated by bad actors.
The crypto space is a survival of the fittest. For ordinary people, our focus should not be on how many multiples we can earn, but rather on how to minimize the chances of being liquidated or manipulated by large players. So how can we reduce that possibility as much as possible?
Examine the chip structure.
The data platform ASXN has organized several Hyperliquid data points that are very instructive.
For example, the OI/market cap ratio data. Open interest (OI) divided by market cap, expressed as a percentage. A higher ratio indicates that derivative positions account for a larger share of the token volume, making it easier to be driven by capital. A typical example is that the positions inherited from the previous HLP liquidation vault once exceeded 40% of JELLY's circulation, causing the token price to tumble like a domino effect.
Similarly, the DEX liquidity table can quickly check the on-chain liquidity of tokens and the risk of manipulation, as well as identify signs of holder accumulation.
The funding rate and comparison with major centralized exchanges allow us to identify potentially manipulated assets. When large positions enter thin open interest, the funding rate may show anomalies compared to other exchanges.
Measure depth.
How to measure depth? How much does it cost to test pushing the price up and down by 2%: How much buy volume is needed to push the price up by 2%. How much sell volume is needed to push the price down by 2%.
This is the real thickness of the best buy/sell orders on the outer side, which is the attack cost for large players. Assets with poor depth = lower costs can distort prices, leading to a lower manipulation threshold and thus higher risks.
This also means that we should trade the most liquid tokens on the most liquid platforms as much as possible.
Taking the core BTC/USDT as an example, within a 0.01% price difference range, the liquidity depths of several currently mainstream perp DEXs are: edgeX $6M, hyperliquid $5M, Aster $4M, Lighter $1M.
This also means that the liquidity of other altcoins will be far below this level and depth, greatly increasing the risk of manipulation, let alone for pre-market trading tokens like XPL.
Be sure to read the liquidation agreement.
Those who can see the rules clearly usually find it easier to understand the risks they can bear.
Before starting formal trading, be sure to read all the liquidation rules: How is the marked price calculated? Does the platform rely more on external index prices, or is its own order book heavily weighted? Especially for pre-market/cold tokens, be wary of whether the Mark is easily influenced by internal transactions. You can monitor the Mark/Index/Last three numbers during normal and volatile periods to see if there's severe deviation.
Is there a tiered margin? If so, the larger the position, the higher the maintenance margin, which means the liquidation point shifts closer. What is the liquidation path—market orders sweeping the order book or building positions in batches? What is the liquidation ratio? What are the trigger conditions for backup liquidation? Are there liquidation fees? Where do liquidation fees go—team income, vault, or foundation buyback?
Additionally, even if two trading platforms hedge for insurance, there is still a very high risk when ignoring depth and liquidation mechanisms. Cross-platform hedging ≠ margin on the same platform, these two things need to be viewed separately. If you can stop loss, you must set a stop loss; if you can use isolated margin, then do so, as isolated margin only affects that position and its margin, while cross margin only allows you to place liquidity that you can afford.
Lastly, it should be noted that the risks of contract mechanisms are greater than most people imagine.
