This article is for educational purposes only. Leverage.Trading is an independent educational and analytics publisher and not a broker, exchange, or investment advisor. Trading with leverage, margin, futures, or derivatives carries a high risk of rapid or total loss. This content is not financial advice and should not be used as a substitute for independent research or professional advice.
When a leveraged position loses enough value, the exchange closes it automatically. No phone call. No warning. No chance to add margin. The system that does this is called a liquidation engine, and it runs on every crypto derivatives platform without the trader seeing it until it fires.
Most traders know the engine exists. Few know how it actually works — what price it watches, when it fires, what happens to the margin, and what happens when the close itself goes wrong.
The liquidation engine is not built to help the trader. It exists to protect the exchange and the traders on the other side of the position. By the time the engine fires, the exchange is managing its own exposure, not the trader’s.
Exchanges run leveraged books. On the other side of every losing trade is a winning trade that expects to be paid. When the losing side runs out of collateral, the exchange is the one left short. The liquidation engine exists to make sure that doesn’t happen.
It monitors every open position in real time. When the equity in an account drops far enough, it closes the position automatically. No human in the loop. No warning call. No grace period. The name on the account doesn’t matter. The math does.
It is not protecting the trader. By the time the engine fires, the exchange is managing its own exposure.
The engine does not watch the chart. It watches mark price, a composite reference built from spot exchange data and funding adjustments, calibrated to track the true underlying price rather than the last individual trade on the futures book.
Mark price and chart price run in parallel until they don’t. The divergence is small in quiet markets. In volatile ones, it can be enough to close a position that looked safe on the chart. That’s not a glitch. That’s how the system is designed.
A candle can wick through a liquidation level on the chart without the engine firing. A position can also be closed without any visible candle touching the level, because mark price crossed the threshold while the chart showed something else. Both happen regularly. Neither is an error.
The chart price is what traders watch. Mark price is what the engine watches. They’re not always the same number.
What most traders think: Liquidation fires when the chart price hits the liquidation level.
What actually happens: Liquidation fires when mark price, the exchange’s composite reference price, reaches the liquidation level. The chart price and mark price diverge during volatile markets, which is exactly when liquidations happen most.
The engine fires when equity falls to the maintenance margin floor. That’s the minimum the exchange requires to keep a position alive, set as a percentage of the position’s notional value.
Take a $100,000 BTC position, 10x leverage, $10,000 in margin. The exchange doesn’t need all of that to stay at risk. It only needs $500. That’s 0.5% of the position size, the maintenance margin floor. When the equity in the account drops to $500, the engine fires. Not at zero. Not at the last dollar. At $500, while $9,500 worth of losses has already happened.
The liquidation price is wherever the market has to move for equity to reach that $500 floor. Below that sits the bankruptcy price, where equity would reach exactly zero. The engine always fires before that point, leaving a gap between the two that the insurance fund is designed to fill.
The $500 isn’t the loss. The $10,000 is. The $500 is the line that activates the engine.
The liquidation price calculator shows exactly where that threshold sits for any combination of entry price, leverage, and position size.
Liquidation fires at $500 equity remaining on a $10,000 margin deposit. The trader does not lose just the $500. They lose the full $10,000. The $500 is the point at which the engine fires, not the amount lost.
The engine doesn’t close a position all at once if it can help it. On exchanges that offer it, the first move is a partial reduction: cut the size enough to bring the margin ratio back above the maintenance floor. This isn’t the exchange doing the trader a favour. It’s the engine trying to avoid the messier outcome of a full close.
When that doesn’t work, or when the exchange doesn’t offer partial liquidation at all, and many don’t, the whole position goes at market price. Whatever the book gives. If that execution price lands below the bankruptcy price, the exchange is short money on the trade. The insurance fund covers it. The fund is built from the small surpluses collected when positions close above the bankruptcy price during normal conditions. The overages from clean liquidations accumulate and wait for the ones that aren’t.
And if the fund can’t cover it? That’s when auto-deleveraging fires. Profitable traders on the opposite side of the market get their winning positions partially closed, without warning, without consent, to cover the deficit. It’s rare. It’s also real. The exchanges with the biggest insurance funds rarely get there. The ones with thin funds do.
ADL fires against profitable traders without warning, on their winning positions, to cover losses from someone else’s liquidation. It is rare but it is real. Exchanges with larger, healthier insurance funds trigger ADL less frequently.
What happens to the margin in all of this is simpler than traders want it to be. When the engine closes a position, the margin deposit is gone. All of it.
In the example above, the engine fires at $500 of equity remaining on a $10,000 deposit. The trader loses $10,000. Not $500. Not just the difference between entry and liquidation price. The full amount deposited into that position.
If the position closes at a price better than the bankruptcy price, the surplus goes to the insurance fund. Not back to the trader.
Most major exchanges offer negative balance protection. The trader cannot owe more than what was deposited. The gap between the liquidation price and the bankruptcy price is the exchange’s problem to solve.
The math ends at the liquidation price for the trader. What happens between that and the bankruptcy price is the exchange absorbing whatever the margin didn’t cover.
Every liquidation is a market order. Market orders move price. And a lower execution price from one liquidation pushes the next position’s equity closer to its own maintenance floor.
Long liquidations push price down. Lower prices reach more long liquidation levels. More market orders follow. The engine isn’t designed to cascade, but the conditions are always present when leveraged positions stack at similar price levels and enough of an initial move reaches the first cluster.
That is the mechanical basis of a liquidation cascade. The engine is doing exactly what it was built to do. The market structure does the rest.
Not all liquidation engines are equal. The differences show up when the market moves fast.
Mark price methodology is the first variable. An exchange drawing its mark price from deep, liquid spot markets is harder to manipulate and resists brief wicks. A poorly constructed calculation produces liquidations that a better-designed system would never trigger. Same position, same market move, different engine, different outcome.
Whether partial liquidation is available is the second. On exchanges that offer it, the engine tries to trim the position first, buying time for a short-duration spike to reverse. On exchanges that skip straight to full closure, the entire position closes the moment the threshold is breached. The difference matters most when the volatility that triggered the close is temporary.
Insurance fund size is the third. A large, healthy fund absorbs shortfalls without reaching ADL. A thin one gets there faster, which means profitable traders on the other side of the market face forced closes on their winning positions more often.
Most major exchanges publish their fund balance publicly, updated in real time. Most retail traders never check it before choosing a platform. The engine runs behind every position they hold, regardless.
A liquidation engine is the automated system a crypto exchange uses to force-close leveraged positions when a trader’s equity falls below the maintenance margin requirement. It protects the exchange and counterparties from losses the trader can no longer cover, and it operates without human intervention.
Mark price triggers the liquidation engine, not the last traded price shown on the chart. Mark price is a composite reference built from spot exchange data, which means it can diverge from the chart price during volatile markets, causing positions to close without any visible candle touching the liquidation level.
The entire margin deposit is lost when a position is liquidated. If the engine fires when $500 of equity remains on a $10,000 deposit, the trader loses the full $10,000, not just the $500 at the threshold. Any surplus from a close above the bankruptcy price goes to the insurance fund, not back to the trader.
The liquidation price is where the engine fires. The bankruptcy price is where the trader’s equity would reach exactly zero. The engine always fires at the liquidation price, which is better for the trader than the bankruptcy price, leaving a gap that the insurance fund is designed to cover in the event of a bad close.
Auto-deleveraging is the last resort in the liquidation waterfall. When the insurance fund cannot cover a shortfall from a forced close, the exchange partially closes positions held by profitable traders on the opposite side of the market to cover the deficit. ADL is not a routine process. It fires only when the insurance fund is insufficient to absorb the loss.
The liquidation engine is the exchange’s risk management system. It closes positions to protect the exchange and its counterparties, and it fires well before losses reach zero.
The sequence runs in one direction. Equity falls to maintenance margin, engine fires, position reduces or closes, insurance fund absorbs any shortfall, ADL handles what the fund cannot. Each step escalates because the one before it didn’t work.
The trigger is mark price, not the chart. The threshold is maintenance margin, not zero. The loss is all deposited margin, not just what remained when the engine fired. Most traders learn all three only after the position is gone.
This article is published under Leverage.Trading’s leverage trading & crypto derivatives education , an independent risk-first learning system built to help traders quantify and manage risk before trading.
