Liquidation is the forced closure of a leveraged position by the exchange when its collateral can no longer cover the loss — and in crypto it is a system event, not a private one. On volatile days, public liquidation feeds show billions of dollars of positions closing in hours: during the global market unwind of August 5, 2024, more than a billion dollars of crypto derivative positions were liquidated in a single day as price gapped through tier after tier of margin.
Bitcoin Trader publishes information, not investment advice. Leveraged trading can result in the loss of all posted margin; this explainer describes clearing mechanics, not strategies.
What triggers a liquidation?
Every leveraged position carries a bankruptcy price — the level at which its posted margin is exactly consumed. Exchanges act earlier, at the maintenance-margin threshold, to keep a buffer. The liquidation engine is automated software that takes ownership of the position and closes it at market the moment the account's effective margin breaches that threshold, charging a liquidation fee designed to discourage running positions to the wire.
Most venues tier the response: partial liquidations that shave position size at successive thresholds before full takeover. The intent is orderly de-risking, but the mechanism has an unavoidable market footprint — the close executes as a market order, and a sufficient mass of simultaneous closes becomes its own price event.
Why do liquidations cascade?
Because forced selling begets forced selling. A cluster of long liquidations pushes price lower, which pushes the next cohort of accounts under their maintenance thresholds, whose liquidations push price lower still. The loop is strongest where leverage is highest and liquidity thinnest — perpetual futures on altcoins — and weakest where books are deep. Exchanges mitigate by marking positions against an index of several venues rather than their own last trade, so a single-venue flash wick does not trigger stops that the wider market never confirmed.
The signature of a cascade is visible in public data: liquidation spikes arrive with volume spikes, funding spikes, and a perp premium blowing out as shorts or longs scramble. Traders who have lived through a few of these treat extreme leverage not as a returns dial but as an expiry date on their own position.
What is the insurance fund for?
Gaps. When price moves so fast that a position closes beyond its bankruptcy price, the loss would leave a negative balance. The insurance fund — a pool accumulated from liquidation fees and excess liquidation proceeds — absorbs that shortfall so it does not land on other traders or the exchange's own capital. Every major derivatives venue publishes its fund's balance, and the balance's health is a real due-diligence item: a thin fund means tail events get handled some other way.
That other way is the mechanism most traders have read about but few have met.
What is auto-deleveraging?
Auto-deleveraging, or ADL, is the exchange's last resort when liquidation losses exceed the insurance fund: the system forcibly closes profitable opposing positions at bankruptcy price, starting with the most profitable and most leveraged counterparties. If a cascade of long liquidations outruns the fund, the shorts who profited from the crash may find portions of their positions closed by the venue itself, at the engine's price, without consent and without appeal.
The queue position is computed from profit and effective leverage — the more a counterparty made and the less margin behind it, the earlier it is deleveraged. ADL is rare on the largest venues outside extreme events and common enough on smaller ones to be a documented feature of their terms. Reading those terms before a crisis is cheaper than reading them during one.
How do traders read liquidation data?
Public liquidation metrics — aggregate notional closed by side over a window — serve as a stress gauge rather than a directional signal. Two disciplined uses stand out. First, positioning context: large clusters of long liquidations indicate the market just removed crowded leverage, mechanically changing who holds risk. Second, market-quality checks: repeated wicks that reverse without follow-through often coincide with localized liquidation bursts, a signature of thin liquidity rather than informed flow.
The undisciplined use is treating liquidation maps as targets — the idea that price 'must' travel to a cluster so the engine can consume it. Clusters are where liquidations would occur if price arrived, not a schedule of where price must go; positioning data describes exposure, never obligation.
Who bears counterparty risk in all this?
Anyone holding a leveraged position bears venue risk on top of market risk: the exchange is the clearing counterparty, the liquidation engine is its enforcement arm, and ADL delegates tail losses to its most profitable customers. That architecture is why institutional flow historically splits between exchange-traded derivatives and cleared or OTC structures, and why regulators, including the U.S. Commodity Futures Trading Commission in its oversight of listed futures, treat margining and default-waterfall design as core protections rather than technicalities.
The practical checklist that follows from the mechanics is short: know your bankruptcy price, not just your liquidation price; treat the insurance fund's published size as part of venue selection; and remember that in the worst case, the exchange's waterfall — liquidation, then fund, then ADL — ends with someone else's position closed by your broker, and vice versa.
What do public liquidation feeds show — and omit?
The aggregate figures quoted after every violent session — billions liquidated — come from data vendors watching exchanges' liquidation feeds and summing them. The coverage is uneven by construction. Some venues publish every forced close in real time; others publish aggregate events with delay, cap the per-event size they report, or omit liquidations entirely — so cross-venue totals are lower bounds, not exact figures, and different vendors' numbers for the same day can differ by hundreds of millions without either being wrong about what it saw.
The omissions matter directionally. Vendor feeds typically count liquidations only above a threshold, so the long tail of small accounts — the retail experience of a cascade — is undercounted relative to the whale prints. Feeds report notional closed, not losses suffered: a 100-million-dollar liquidation may have destroyed only the last few million of margin, so the figure prices forced activity, not pain. And the feeds run only while the venue's systems do — during the outages that accompany the worst cascades, the missing minutes are precisely the most liquidation-dense. The disciplined reading treats liquidation data the way it treats any vendor aggregate: directionally true, definitionally noisy, and most useful in comparing events across time on the same vendor's methodology rather than as an absolute measure of destruction.
For more context, read How Leverage and Margin Trading Work on Crypto Exchanges.
For more context, read what is open interest.
For more context, read How the Funding Rate Basis Trade Works.
