Auto-Deleveraging (ADL)

A detailed look at Auto-Deleveraging (ADL) — the last-resort mechanism BULK uses when a liquidation alone can't fully resolve a bad position. For a gentler introduction, see the ADL section of How Liquidations Work; for the formal specification, see docs.bulk.trade/bulk-exchange/ADL.

The mental model: four layers before ADL ever gets involved

It helps to think of BULK's risk system as a stack of defenses, where ADL is genuinely the last resort, not the first response to a losing position:

Impact-aware margin — your maintenance margin requirement already accounts for how expensive it would be to close your position, before anything goes wrong. See Understanding Portfolio Margin.
The portfolio optimizer — when a position does need to shrink, BULK calculates which specific reduction restores your margin most efficiently, rather than closing indiscriminately. See How Liquidations Work.
Protected execution — the actual reduction happens within deterministic price limits anchored to fair and oracle prices, using available market liquidity.
Risk vault and ADL — only if the first three layers can't fully resolve the situation does the system reach for ADL (or, in the worst case, the protocol's own risk vault absorbs the remainder).

Most liquidations never reach step 4 at all. ADL exists specifically for the situations the first three layers can't fully cover.

The two separate triggers

ADL activates for two genuinely different reasons, and it's worth keeping them distinct because they behave differently:

Shortfall: The liquidated account's realized losses exceed its remaining collateral. There's a real hole to fill.

Underfill: Some portion of a planned liquidation simply couldn't be filled within the protocol's capacity — for instance, if the market doesn't have enough liquidity to absorb the reduction. Crucially, underfill can trigger ADL even when the liquidated account has no collateral shortfall at all — it's a capacity problem, not necessarily a solvency problem.

How BULK picks who gets deleveraged

ADL doesn't pick targets randomly, and it doesn't just grab the largest positions. Eligible accounts are the ones holding profitable positions on the side of the market needed to offset the shortfall or underfill — an account with no position in the right direction, or one that's currently at a loss, isn't a candidate at all.

Among eligible accounts, BULK ranks by a score that blends two things: how much profit the account is sitting on, and how much leverage it's using, using a configurable weighting between the two. In plain terms — the accounts most likely to be selected are the ones with the largest combination of "large profitable position" and "high leverage on that position." This isn't arbitrary: it targets the accounts best positioned to absorb a reduction (large realized profit cushion) while also weighting toward accounts whose leverage suggests they're actively seeking outsized exposure, rather than deleveraging small, conservatively-sized winning positions first.

Accounts already being liquidated, and protocol-owned accounts, are explicitly excluded from ever being selected as ADL candidates.

How the reduction is actually applied

Allocation is spread pro-rata across eligible accounts by their score — higher-ranked accounts absorb a larger share.
No account can be pushed through zero — ADL can shrink a position, but it will never flip a long into a short (or vice versa) to satisfy an allocation.
Very small allocations are deliberately concentrated on the highest-ranked accounts rather than being spread as negligible "dust" across many accounts — nobody gets an ADL notice for a rounding error.

Execution price depends on why ADL fired

This is a subtle but important distinction:

Underfill ADL reduces the selected positions at the current mark price — the same reference price used everywhere else on BULK.
Shortfall ADL closes the selected profitable positions at their own entry price, while the offsetting reduction on the other side is recorded at mark price. The difference between those two prices is exactly what covers the shortfall. This means a shortfall-driven ADL event effectively transfers the covered loss to the deleveraged account at a price that's more favorable to them than an underfill-driven one — a deliberate design choice, not an accident.

Every ADL event reports its executed price, size, side, and the reason (shortfall vs. underfill) it fired.

The risk vault: the actual last resort

For shortfall events, a configured share (which can vary by market) may be routed directly to BULK's protocol risk vault rather than to ADL candidates at all. ADL covers the remaining eligible amount — and if even ADL's candidate pool doesn't have enough capacity to fully cover the shortfall, the vault absorbs whatever's left. For underfill events, any quantity that can't be allocated across eligible accounts is routed back into protected execution rather than simply being dropped or left unresolved.

In other words: even in the genuine worst case, there's a defined, deterministic place for the residual risk to go. Nothing is left as an unaccounted-for loss sitting on the exchange's books.

Why this matters if you're holding a profitable, leveraged position

Being ADL'd isn't a punishment or a sign you did anything wrong — it's the flip side of holding a large, leveraged, profitable position during someone else's liquidation event. If you want to reduce your exposure to being selected:

Lower leverage reduces your ADL score's leverage component.
There's no way to opt out entirely if your position genuinely is the best-suited counterparty — that's the mechanism working as intended, not a bug to route around.