Crypto auto-deleveraging: what remains after ADL?

Crypto auto-deleveraging (ADL) is an exchange risk mechanism that can forcibly reduce an opposing derivatives position when the venue's loss-management conditions are met. A position can be profitable and still be selected. After ADL, verify the actual settlement, remaining position and working orders before treating a hedge or automated strategy as intact.

This AI-assisted guide uses official Bybit and Hyperliquid documentation reviewed on 23 October 2026, plus a hypothetical hedge ledger. Its purpose is to explain an involuntary position change and the decisions that follow; it does not predict an ADL event or recommend a hedge.

ADL and your own liquidation are different events

A trader's liquidation concerns that account or position failing the venue's margin requirements. ADL can affect the opposing side, including a trader whose position has gained value. Your liquidation price therefore cannot answer every question about whether your position will remain open. Our mark-price and liquidation guide covers that separate margin mechanism.

Bybit describes selecting opposing positions using an ADL ranking based on leveraged returns, with profitable positions prioritized; losing opposing positions may also be selected. Its documented settlement uses a bankruptcy price defined by the insurance fund. These are Bybit perpetual and expiry contract rules, not universal contract terms. Bybit's ADL mechanism explanation provides the applicable selection and settlement details.

Do not assume every venue waits for an insurance balance to reach zero. Bybit's current API documentation includes both contract-PnL drawdown triggers and insurance-pool equity triggers. A positive displayed balance alone does not establish that ADL cannot occur. Product parameters and pool membership matter. Bybit's ADL alert documentation explains these conditions.

A queue indicator is not an event probability

An ADL indicator describes a position's place in the venue's selection system. It does not, by itself, supply a calibrated probability that an event will occur within the next hour. Reading a relative ranking as a forecast omits whether the loss-management trigger will be reached and how much exposure the venue will need to reduce.

Keep three questions separate:

• What conditions activate ADL for this product?
• If activated, how does the venue rank and select positions?
• At what price and quantity will an affected position settle?

The answers can differ across venues. Hyperliquid documents ADL when account value or isolated-position value becomes negative, selection using unrealized PnL and leverage, and closure against the underwater user at the previous mark price. That differs from Bybit's stated bankruptcy-price settlement. Hyperliquid's official ADL documentation is the source for that scope.

Do not copy a queue formula from one exchange into another exchange's monitoring rule. Even a correct historical formula needs a dated product reference and the correct account inputs.

Worked example: the profit survives, but the hedge shrinks

Consider a hypothetical trader with 0.100 BTC in spot at one venue and an isolated-margin USDT-linear short representing 0.100 BTC at another. Both entry references are 60,000 USDT/BTC. For arithmetic only, assume spot and derivative reference prices move together, USDT maintains its assumed value, and fees, funding, basis and collateral changes are absent.

The linear payoff is deliberate. The linear versus inverse contract comparison explains why an inverse contract would require reciprocal-price PnL and a different settlement-balance ledger.

At 52,000, assume an ADL event closes 0.060 BTC of the short at exactly 52,000. This settlement price is an invented input, not a prediction of either exchange's ADL price. The spot holding remains untouched at the separate venue.

The closed short's gross realized price profit is 0.060 × (60,000 − 52,000) = 480 USDT. The remaining short has 0.040 × (60,000 − 52,000) = 320 USDT of gross unrealized profit. The spot has an unrealized loss of 0.100 × (52,000 − 60,000) = −800 USDT.

At this deliberately frictionless observation, 480 + 320 − 800 = 0. ADL has changed the split between realized and unrealized results, and it has changed the future exposure. A positive realized-profit entry is not evidence that the original hedge is still in place.

Now assume both prices fall from 52,000 to 50,000 with no further trades. Spot loses another 200 USDT; the smaller short gains 80 USDT. The combined change is −200 + 80 = −120 USDT, also calculated as 0.060 × (50,000 − 52,000).

A rise of 2,000 would instead produce a positive 120 USDT change under the same assumptions. Neither direction is the point: the post-event portfolio has become directionally long. Real positions add funding, execution costs, basis and settlement differences; the funding-cost ledger helps keep those cash flows visible.

Reconcile the event before repairing the strategy

Use a small incident worksheet rather than relying on the notification alone:

Bybit identifies ADL in its API enums with AdlTrade and CreateByAdl_PassThrough. These labels can help distinguish a system-driven reduction from an ordinary strategy exit, but their presence does not replace the quantity and cash ledger. Bybit's execution and order enums define those values.

Its position API exposes position size, adlRankIndicator and position update information. Read the documented product scope and pagination rules when collecting account state; a response covering one symbol is not a complete portfolio inventory. Bybit's position information API documents those fields.

For an exit that remains active, quantity is only part of the state. The trailing-stop ledger shows how tracking start, saved extreme and trigger threshold also matter. Reconcile that state with the remaining position before treating an old trailing instruction as appropriate coverage.

Decide explicitly whether to pause new strategy instructions while the state is unresolved. A pause is not proof that accepted orders have stopped. The kill-switch guide explains why new-work controls and exchange exposure need separate verification.

The failure case: automatic replacement recreates the exposure

Suppose the hedge controller still expects a 0.100 BTC short and immediately sells the missing 0.060 BTC after the event. It may restore the original quantity, but it also submits a new trade during the conditions that produced ADL. A second component may simultaneously submit the same repair after reading the old position snapshot.

The result can exceed the intended short. Before placing a repair, account for pending requests and fills, then make one explicit decision about the target exposure using current records. If a request times out, do not treat silence as a failed trade; follow the order-timeout reconciliation process.

Bybit's ADL help page also says affected traders' active orders are canceled. That is a reason to inspect protection after the event, not a reason to assume another venue behaves identically or that every local conditional instruction has disappeared. Verify the actual order state and any automation that can recreate orders.

Write the ADL response before opening the position

A useful plan names the product's current rule source, the exposure to review after a reduction, who may authorize a replacement, and the evidence needed before automation resumes. Record how a partial reduction changes the portfolio, rather than only whether the closed leg booked a profit.

Backtests and paper environments may omit ADL entirely. A useful scenario test can inject an explicitly hypothetical forced reduction and ask whether the controller detects the mismatch, inventories pending work and stops duplicate repair attempts. It tests recovery behavior, not the probability of a real ADL event or the adequacy of an exchange's insurance resources.