Liquidation cascade

A liquidation cascade is a chain reaction in leveraged markets where forced closures of positions move the price far enough to trigger the next round of liquidations. It turns a modest price move into a much larger one, in either direction.

How it works

Every leveraged position has a liquidation price. When the loss on a position eats through the margin until it falls below the exchange's maintenance requirement, the liquidation engine closes the position with a market order. A long is closed by selling, a short by buying. Nobody decides anything at that point; a rule is executed.

A liquidation cascade starts when those forced orders are large relative to the order book. The forced sell from one liquidated long pushes price down, which brings the next long to its liquidation price, which produces another forced sell. The output of each step is the input of the next.

Liquidation prices are not spread evenly. Traders enter near the same levels (round numbers, breakouts, recent highs and lows) and use similar leverage, so their liquidation prices land in narrow bands. A rough approximation for an isolated long is:

Liquidation price ≈ entry × (1 − 1/leverage + maintenance margin rate)

With a $100,000 entry and a 0.5% maintenance rate, that puts 25x longs near $96,500, 20x longs near $95,500 and 10x longs near $90,500. Exact numbers depend on the exchange's tiers and fees, but the clustering is the point.

How far a cascade runs depends on three things: how much leveraged exposure sits in the bands, how close together the bands are, and how much resting liquidity there is to absorb the forced orders. The mechanism runs both ways. When shorts are crowded, a rally liquidates shorts, their forced buys lift price into the next band of shorts, and the result is a short squeeze.

DeFi lending has its own version. When collateral on a protocol such as Aave falls below its liquidation threshold, liquidators repay part of the debt and take the collateral at a discount, then usually sell it. If many protocols accept the same collateral, read the same oracle and liquidate into the same pools, one forced sale lowers the price every other protocol sees on the same block. Borrowed assets that are redeposited as collateral elsewhere link protocols that never reference each other.

A worked example

BTC trades at $100,000 after a rally. Long positions opened near that price sit in three bands:

Leverage Approx. liquidation price Positions
25x $96,500 $30 million
20x $95,500 $60 million
10x $90,500 $100 million

A piece of news pushes price down 3.5% to $96,500. To keep the arithmetic simple, assume every $20 million of market selling moves price down another 1%.

  • The 25x band liquidates: $30 million of forced selling, −1.5%. Price: $96,500 × 0.985 = $95,052
  • That is below $95,500, so the 20x band liquidates: $60 million, −3%. Price: $95,052 × 0.97 = $92,201
  • $92,201 is still above $90,500, so the 10x band survives and the selling stops.

Total move: −7.8%, of which 3.5% came from the news and about 4.3% from forced selling.

Now halve the book depth, so $10 million moves price 1%. The 25x band takes price to $93,605, the 20x band to $87,989, which is below $90,500, and the 10x band's $100 million pushes it to $79,190. The same news and the same positions now produce a −20.8% move. Book depth decides whether a cascade fizzles or runs.

What it means in practice

Exchanges publish forced orders as they happen, split into longs and shorts. Liquidations every few minutes are normal. What stands out is a sudden burst concentrated on one side.

A drop driven mainly by liquidations behaves differently from one driven by a change in fundamentals. Once the clustered leverage is cleared, the forced orders stop arriving as suddenly as they started, which is why cascades tend to end in a sharp wick and a partial snap-back rather than a slow fade. A repricing on new information tends to keep going. Spot holders feel cascades too: their position did not cause the move, but they absorb the price.

The fuel is usually visible beforehand. Rising open interest, persistently positive funding and a long stretch of low realized volatility together describe a market where a lot of leverage has been added at similar prices and nobody has been forced out for a while.

When it matters most

Cascades are most likely when open interest is high relative to spot liquidity and the book is thin: weekends, quiet sessions, and the period right after a calm stretch when traders have tightened stops and raised leverage. Crypto combines high available leverage, 24/7 trading and thinner books than most equity markets, which is a large part of why cascades are more common here.

They also matter around obvious levels. A break below a well-watched support level triggers stop-losses and liquidations at the same time, which produces the sharp wick and reversal described under liquidity sweep.

Common misconceptions

"A crash like that must be panic selling." Panic can add to it, but the first leg of a cascade is usually automated. Liquidation engines execute regardless of anyone's view, and they do it faster than people can react.

"Low leverage on my own position keeps me out of it." Your position does not need to be liquidated for you to be hit. A spot holder or a conservatively collateralized DeFi borrower still marks to the price that other people's forced selling creates.

"Cascades only go down." Short squeezes are the same mechanism in reverse. Crowded shorts and negative funding can produce a cascade upward.

"A conservative loan-to-value ratio means a DeFi protocol is safe." The ratio covers one protocol. It says nothing about how many times the same collateral has been borrowed against elsewhere, or about the depth of the pool all those protocols liquidate into.