How a Liquidation Cascade Builds, Step by Step

A liquidation cascade is a chain reaction in which forced sell-offs drive a price down, and that lower price triggers more forced sell-offs, until the selling burns out or the market runs out of positions to close. It happens when many traders have borrowed money to bet on a rising price, the price falls to the point where their collateral no longer covers the loan, and an automated system sells them out one after another (each sale pushing the price a little lower and dragging the next trader over the edge).

The counter-intuitive part is that the drop does not need a big piece of news to start. A single large sale in a thin market can be enough to knock the first row of over-borrowed traders offside, and from there the machinery does the rest. The trigger is often small. The result rarely is.

Key takeaways

• A liquidation cascade is a feedback loop: forced sales lower the price, the lower price forces more sales. Leverage (borrowed money used to size up a bet) is the fuel.
• The forced sales are automatic. When a trader's collateral falls below a set threshold, an exchange or protocol closes the position without asking, to prevent the trader from owing more than they put in.
• Cascades are worst in thin, low-liquidity conditions, because each forced sale moves the price more, which trips the next one faster.
• Guardrails exist: insurance funds, auto-deleveraging, partial liquidations, and circuit breakers all aim to interrupt the loop before it runs to the bottom.
• Cascades are visible after the fact because every liquidation is recorded onchain or on-exchange, which is why researchers and newsrooms can reconstruct exactly what happened.

What is being liquidated, and why it happens without warning

Most of this activity happens in a market called perpetual futures, or "perps": contracts that let a trader bet on the price of an asset without owning it, with no expiry date. Crucially, they let a trader post a small amount of money (the margin, or collateral) to control a much larger position. That is leverage.

Think of it like a mortgage on a house. You put down 10 percent and borrow the rest. If the house rises 10 percent, you double your deposit. If it falls 10 percent, your deposit is gone. The difference in these markets is speed and automation. When your collateral is exhausted, there is no grace period and no phone call. An automated system closes your position at the market price to cap the loss at the margin you posted. That cap is not a guarantee: in fast markets the position can close below the liquidation price, and venues cover the shortfall from an insurance fund or, if that is exhausted, by socialising it across profitable traders. That forced close is a liquidation.

Here is why one liquidation can beget the next. Closing a long position (a bet that the price rises) means selling. Selling pushes the price down. If enough traders are clustered at similar liquidation prices, the sale that closes one trader drops the price into the next trader's liquidation zone, which triggers another sale, and so on. That is the cascade.

A worked example of the loop

Suppose an asset trades at $100 and three traders are long with different amounts of leverage. Leverage of 10x means a roughly 10 percent adverse move wipes the collateral; 20x means roughly 5 percent; 50x means roughly 2 percent. The exact liquidation price depends on fees and maintenance margin, but the clustering is the point.

A single sell order nudges the price from $100 to $98. Trader A is liquidated, and the exchange sells $5m of the asset into the market. In a thin order book, that sale pushes the price to $96. Trader B is now underwater and liquidated, adding another $3m of selling. The price slides to $90, and Trader C follows. What began as a $2 dip became a 10 percent drop, entirely because the sales were forced and stacked. Nothing changed about the asset's fundamentals in those seconds. This illustrative example uses round numbers, not observed data.

This is also why cascades compound faster in thin markets. If the order book were deep enough to absorb Trader A's $5m without moving much, Trader B might never be touched. Liquidity is the shock absorber, and when it thins out, the chain reaction accelerates.

The guardrails, and where they fail

Exchanges and lending protocols know this loop exists and build brakes for it. The common ones:

• Partial liquidation. Instead of closing an entire position at once, the system closes only enough to restore a safe margin, which reduces the size of each forced sale.
• Insurance funds. A pooled reserve that absorbs losses when a position is closed at a price worse than its liquidation price, so the shortfall does not fall on other traders.
• Auto-deleveraging (ADL). When the insurance fund is depleted, some venues automatically reduce the winning traders on the other side to balance the book, a controversial last resort.
• Circuit breakers and price bands. Limits on how far or how fast a price can move before trading pauses, borrowed from traditional exchanges.

These brakes work until they do not. In the sharpest events, forced sales outrun the insurance fund and liquidity dries up at exactly the moment it is needed most.

Reading a cascade after it happens

Every forced close leaves a record. On centralized exchanges it sits in their trade logs; on onchain venues it is a public transaction anyone can read. That is why a cascade can be reconstructed second by second: which positions closed, at what price, in what order, and how much selling each one added.

The difficulty is comparison. To measure a cascade properly, or to compare one across venues, each liquidation event has to resolve to the same fields: the asset, the trader address, the position size, the liquidation price, the timestamp, and the realized loss. Raw data from different chains and different protocols encodes these things in incompatible ways, so a liquidation on one venue and a liquidation on another are not directly comparable until they are normalized to a shared schema. Allium provides a normalized read layer that standardizes these records across many blockchains into consistent fields, which is what lets researchers and newsrooms study cascades rather than just witness them.

Allium is a data infrastructure layer, a way to observe what occurred. It is not an exchange, a broker, or a market maker, and it does not offer trading or investment advice. The same normalized records serve research as readily as reporting, from academic study of onchain derivatives to newsroom reconstruction of market events.

What is still unsettled

Two questions do not have clean answers. First, how much leverage a venue should permit is a live debate: higher limits attract volume but tighten the clustering that makes cascades severe. Second, the regulatory treatment of high-leverage crypto derivatives varies sharply by jurisdiction, and venues registered in different jurisdictions sit under different frameworks from regulated futures exchanges. That gap is a matter of active policy discussion rather than settled law, and it shapes how much of this activity happens where, and under whose rules.

Frequently asked questions

What is the difference between a liquidation and a liquidation cascade?

A liquidation is a single forced close of one over-leveraged position. A liquidation cascade is many liquidations chained together, where each forced sale moves the price enough to trigger the next. One is an event; the other is a self-reinforcing loop.

Can a liquidation cascade happen without any real news?

Yes. Because forced sales themselves push the price, a cascade can begin from a single large order in a thin market, with no change to the asset's fundamentals. The clustering of liquidation prices and low liquidity are enough to start and sustain the loop.

Do liquidation cascades only happen in crypto?

The mechanism (leverage plus forced selling plus feedback) exists in any leveraged market, including traditional futures and margin equities. It is more visible and often faster in crypto because leverage limits can be high, markets trade around the clock, and liquidations are automated with no human discretion.

What stops a cascade once it starts?

Selling stops when the over-leveraged positions in that price range are exhausted, when fresh buyers step in at lower prices, or when a venue's brakes engage: partial liquidations, insurance funds, auto-deleveraging, or a circuit breaker that pauses trading.

Are short positions liquidated in a cascade too?

Yes, in the opposite direction. A cascade of short liquidations (bets that the price falls) forces buying, which pushes the price up and triggers more short liquidations. These upward cascades are sometimes called short squeezes.

How can researchers measure a cascade after it happens?

Every liquidation is recorded, onchain or in exchange logs. Reconstructing the sequence requires normalizing each event to shared fields such as asset, size, price, timestamp, and loss so events across different venues and chains can be compared consistently.

Interested in learning more about Allium’s onchain data infrastructure? Speak to someone on the team.

Allium provides onchain data infrastructure. Companies named in this article may be Allium customers, prospects or commercial counterparties. This article is informational only and is not investment, legal or tax advice. Data and information last reviewed: September 23, 2026.