Skip to content
Financial Context Wiki

Auto-Deleveraging (ADL)

Auto-deleveraging forcibly reduces selected opposing derivatives positions under a venue-specific loss waterfall; trigger, ranking, price, quantity and fees are not universal.

Updated

For education only. ADL can forcibly reduce a profitable or hedging position under venue-specific rules; verify the applicable entity, product, fund, queue, match price and account ledger.

Direct answer

Auto-deleveraging (ADL) is a derivatives venue’s forced reduction of selected positions opposite distressed or bankrupted positions. It is usually a late stage in a loss waterfall after ordinary liquidation or other backstop capacity is inadequate. The selected trader can have a profitable position forcibly closed in whole or in part at a venue-defined price. ADL is not ordinary liquidation of that trader, an insurance payout or necessarily an on-chain mechanism.

There is no universal trigger, queue formula, match price or fee rule. Bybit documents insurance-pool drawdown conditions, leveraged-return ranking and bankruptcy-price matching. OKX documents security-fund thresholds, leverage-PnL ranking and normally mark-price matching, with possible bankruptcy-price use near fund depletion. Hyperliquid documents negative account or isolated-position value and its own ranking and price. Coinbase documents a waterfall that may include partial liquidation, a liquidity support program, insurance, ADL and clawbacks. Each claim is product, entity, region and rule-version specific.

The economic loss is not simply “profit confiscation.” Forced realization can remove future upside, break a hedge, cancel orders, release or reshape margin and require re-entry at a worse basis. A public insurance-fund balance is neither user insurance nor proof that real-time capacity is sufficient.

How it works

  1. Pin the venue and legal entity, jurisdiction, product and symbol, linear, inverse or quanto contract, multiplier, settlement asset, margin and position mode, risk tier, and rule or API version at the event time.
  2. Freeze the pre-event ledger: signed quantity, entry, index, mark and last prices, notional, collateral, equity, initial and maintenance margin, funding, fees, open orders and hedge or portfolio offsets.
  3. Reconstruct that venue’s waterfall from liquidation trigger through partial liquidation, order-book execution, backstop or liquidity provider, insurance or security fund, ADL, clawback or socialized loss.
  4. Calculate the bankruptcy shortfall and applicable fund-pool scope, balance and trigger or stop condition. Do not replace the actual rule with “the fund reached zero.”
  5. Recompute the venue’s current ranking for the applicable product and margin mode. Treat queue lights or percentiles as dynamic snapshots, not occurrence probabilities or execution promises.
  6. Apply the actual matched quantity, venue-defined price, multiplier, rounding and fee rule; reconcile realized PnL, remaining size and margin, canceled orders, released collateral and broken hedge.
  7. Stress repeated ADL, price gaps, funding, re-entry basis, API failure and rule changes; set position, leverage, margin, alternate-hedge and venue-exit limits, then retain notices, fills and rule evidence.

Keep liquidation price, bankruptcy price, mark price, index price, last price and actual execution or ADL match price separate. Linear, inverse and quanto contracts can express notional, PnL and collateral in different units. A reduce-only order or stop can lower exposure but cannot veto a venue risk engine.

Example

  • Venue-specific drawdown. In Bybit’s documented example, an 8-hour insurance-pool high of 20,000 USDC, open-position margin of 1,000 USDC and losses of 8,000 USDC produce (8,000 - 1,000) / 20,000 = 35%. That exceeds the stated 30% trigger; restoring the drawdown to 30% requires 8,000 - 20,000 x 30% = 2,000 USDC. This is a dated Bybit rule example, not an industry formula.
  • Partial queue allocation. Suppose a venue ranks opposing positions A, B and C in that order, with 5,500, 2,500 and 2,000 contracts. If 7,000 contracts must be reduced, A loses all 5,500, B loses 1,500 and retains 1,000, while C is untouched. Ranking and remaining exposure must then be recalculated.
  • Forced-close ledger. A linear USDT short has 2 BTC at entry 70,000; ADL reduces 0.75 BTC at 60,000. Realized PnL is (70,000 - 60,000) x 0.75 = 7,500 USDT, leaving 1.25 BTC. If executable re-entry is 58,000, the closed portion forgoes (60,000 - 58,000) x 0.75 = 1,500 USDT of further short profit before fees, slippage and funding.
  • Contract units. A linear long of 2 BTC from 30,000 to 31,000 earns 2 x (31,000 - 30,000) = 2,000 USDC. An inverse 60,000 USD position over the same move earns 60,000 x (1 / 30,000 - 1 / 31,000) = 0.0645161290 BTC. Similar dollar exposure does not make settlement units or queue inputs interchangeable.

Risks

  • Wrong venue, entity, jurisdiction or rule version.
  • Product, symbol or settlement asset is inapplicable.
  • Linear, inverse or quanto multiplier and PnL units are mixed.
  • Mark, index, last, liquidation, bankruptcy and match prices are conflated.
  • Liquidation waterfall or backstop order is misread.
  • Insurance or security fund pool scope is wrong.
  • Published fund balance is delayed or unavailable.
  • Trigger or stop threshold is hidden, dynamic or changed.
  • Ranking formula or margin mode is misapplied.
  • Queue light or percentile is mistaken for probability.
  • Partial ADL changes remaining size, margin and liquidation risk.
  • Cross or portfolio collateral transmits account-wide stress.
  • One hedge leg is reduced, exposing delta or basis.
  • Active orders are canceled or reduce-only state changes.
  • Re-entry spread, slippage, fees and basis are omitted.
  • Funding settles across the event unexpectedly.
  • Thin liquidity, gap or mark-oracle failure amplifies losses.
  • Repeated ADL compounds a liquidation cascade.
  • API, UI, notification or evidence is late or incomplete.
  • Venue custody, solvency, governance, legal, tax or disaster recovery fails.

Common misconceptions

  • ADL is ordinary liquidation. Liquidation handles the distressed position first; ADL forcibly reduces selected opposing positions at another stage of the waterfall.
  • ADL starts only when an insurance fund reaches zero. Venues use different drawdown, threshold, negative-equity and capacity rules.
  • The highest profit percentage always goes first. Ranking depends on venue, product, margin mode, leverage or maintenance metrics, hedge treatment and live data.
  • ADL takes all profit or is identical to a clawback. It normally realizes PnL on a specified quantity at a rule-defined price; clawbacks and socialized loss are separate mechanisms.
  • A stop, lower leverage or queue light prevents ADL. These may reduce exposure or rank, but they do not guarantee exemption during extreme conditions or rule changes.

Sources

Navigation

Search the wiki...