What Is Hyperliquid ADL? Why Winning Positions Were Force-Closed on October 10, 2025
Columns · · Satoshi Note Editorial
Read in: 한국어
ADL (Auto-Deleveraging) is the last-resort safeguard in which an exchange force-closes profitable positions on the opposite side when liquidations can't cover a loss.
It means a position can be closed even though its owner did nothing wrong.
On October 10, 2025, Hyperliquid's ADL ran at scale.
One paper counted about $2.1 billion ($2,103,111,431) in observed liquidations between 21:16 and 21:27 UTC — roughly 11 minutes.
ADL fired repeatedly in that window.
The episode is still debated: whether ADL was needed, and whether the method was right, depends on which study you read.
Knowing what ADL is and the order it runs in answers "why was my winning position closed?", and it separates what is settled about October 10 from what is disputed.
1. When does ADL run?
A leveraged exchange acts in stages when a position turns dangerous. Simplified:
- Ordinary liquidation: if margin runs short, the position is sent to the order book and closed.
- Backstop: if the book can't absorb it, Hyperliquid's HLP liquidity vault can take it.
- ADL: if the loss still isn't covered — the account value goes negative — the exchange reduces positions on the opposite side to close the hole.
ADL is not an insurance fund; it is settlement among opposite-side users. Hyperliquid's official documentation says opposite-side users are ranked by unrealized PnL and leverage used.
Per third-party explainers, positions are closed at the prior mark price, not the live order-book price.
So the positions ADL reaches are typically those with large profit and high leverage — the side that was most right is reduced first.
Sources describe the exact ranking formula in slightly different ways. This article doesn't compute the queue; it stays with the two inputs the official documentation names.
2. What happened on October 10, 2025
Reports said it was the first time in over two years of operation that Hyperliquid triggered ADL on cross-margin accounts.
What the paper's count shows:
| Item | Detail |
|---|---|
| Date | October 10, 2025 |
| Window | 21:16–21:27 UTC (about 11 minutes) |
| Observed liquidations | about $2.1 billion |
| ADL | fired repeatedly within the window |
The $2.1 billion is the paper's count of observed liquidations. The same figure appears elsewhere to mean realized PnL on the winning side, so check which meaning a source uses before comparing.
Within roughly ten minutes, a great deal of money moved at once, and analyses described liquidations triggering further liquidations in a cascade.
3. What is disputed
The central question is one: was ADL applied only as much as necessary?
- One report argued that the production ADL ordering was outdated and caused about $650 million in excess losses.
- A later academic analysis replayed public data and estimated Hyperliquid's approach overshot by about $45–52 million.
The figures rest on different assumptions and can't replace each other. Nothing here lets us call either one wrong.
They share one point: neither says ADL was unnecessary. The disagreement is over how much, and in what order, positions should have been closed.
4. What it means for a position
ADL means profit does not make a position safe.
- The larger the profit and the higher the leverage, the earlier in the queue.
- Once closed, the position is gone regardless of where price goes next.
- It closes at the prior mark price; you don't choose the fill.
This article doesn't advise how to avoid it. The ordering is exchange policy and can change.
5. What do top Hyperliquid BTC positions look like now?
The two ADL inputs — unrealized PnL and leverage — are visible per account on the Satoshi Index.
At 07:14 KST on November 4, 2026, the five largest top-account BTC shorts were all on Hyperliquid, at 5x, 5x, 25x, 8x and 20x. The 25x short was about $2.91M in profit.
This sample can't predict ADL: the ordering is internal to the exchange and isn't computed on the page.
What you can take from it is the habit of looking at profit and leverage together, not size alone.
Wrap-up: ADL is the last settlement after liquidation
ADL reduces profitable opposite-side positions when liquidation and the backstop can't cover a loss.
On October 10, 2025, about $2.1 billion in liquidations was observed over about 11 minutes, and ADL fired repeatedly.
How much was excessive is disputed: about $650 million in one report, $45–52 million in another.
ADL isn't only the losing side's problem — large-profit, high-leverage positions can be closed too, and that is the core of the mechanism.
Frequently Asked Questions
What does ADL stand for, and how is it different from liquidation?
Auto-Deleveraging. Liquidation closes a losing position. ADL closes profitable positions on the opposite side when that liquidation still doesn't cover the loss. The side that gets closed is the reverse.
If I'm in profit and get ADL'd, do I lose money?
According to third-party explainers, the close happens at the prior mark price, so profit already earned is realized at that price. The cost of ADL is that you didn't choose when or at what price to close, and you can't carry the position forward. Check your own trade history and the exchange documentation for actual fills.
Was the October 10, 2025 ADL a mistake?
It can't be stated flatly. One report argues excess losses; a later analysis puts the overshoot much lower, and the numbers depend on assumptions. Satoshi Note does not take a side.
Check it yourself on the Satoshi Index
On the coin pages of the Satoshi Index, each top account's position size, leverage and unrealized PnL are listed. It opens without login and refreshes every 30 minutes (every 3 hours for dYdX).
Sources and basis: liquidation size and timing from the arXiv paper "Autodeleveraging: Impossibilities and Optimization" (2512.01112); the dispute from a later paper (arXiv 2602.15182) and Protos reporting; ADL ordering from a summary of Hyperliquid's official documentation. Current positions are a snapshot at 07:14 KST on November 4, 2026. This content is market-data analysis, not investment advice.