What is the auto-deleverage mechanism? 

When a user’s position is being liquidated, the user’s remaining position will be taken over by the system. If the position cannot be executed at the liquidation price in the market (this means there are not enough orders on the book) and the insurance fund is insufficient to cover the loss from this position, the auto-deleverage (ADL) mechanism will kick in. ADL will deleverage (close out users’ positions) users with positions on the opposite side, the deleverage queue is based on leverage and PnL.

Detailed queuing formula:


= PNL Percentage * Effective Leverage (if PNL percentage > 0)

= PNL Percentage / Effective Leverage (if PNL percentage < 0)


PNL Percentage = (Unrealized profit and loss)/(Cumulative Open Value +/- Margin Added/Removed)

Cumulative Open Value = Amount of Contracts / Average Open Price

Difference between ADL and clawback?

The Clawback mechanism distributes the loss to all users while maintaining profits on the platform according to position sizes. This mechanism occurs after settlement and transfers part of the user’s profit to share the losses; therefore, the profits of the position cannot be transferred out before settlement.

The auto-deleverage mechanism will prioritize deleveraging users with high leverage and high profits. Since this mechanism is realized by closing out positions, the profits can be transferred out at any time.


Suppose now we have 6 users’ position information as follows:


If now a catastrophic loss occurs on the long side, the order cannot be executed with the pre-designated liquidation price, and the insurance fund cannot cover the loss, we will rank the short side based on Rank Value. The result is: C, B, and F.

Suppose the liquidation position size is 52 and the liquidation price is the current price (9627.5). At this moment, user C’s positions will all be closed, user B will automatically close out 2, and the closing price will be the current price. The two users will be notified that their positions have been closed out. Therefore, users can set up a handling mechanism to reopen the position if such an event happens.

Did this answer your question?