Uncover the crazy truth behind the ADL mechanism. Don’t let the exchange take your money easily

Editor's Note: This article explains the Automatic Deleveraging (ADL) mechanism in the cryptocurrency market, particularly illustrating how ADL compensates for insufficient liquidations by forcibly closing profitable customers' positions. The article mentions that the execution of ADL involves many opaque aspects, such as price discrepancies and execution delays. The author points out that understanding liquidation and ADL mechanisms is crucial for comprehending the crypto market and encourages readers to provide feedback to improve the explanation.

Below is the original content (slightly edited for readability):

Today, I want to write a summary about the Automatic Deleveraging (ADL) mechanism in the crypto market. I'm writing this article because I've recently received many messages about ADL and issues with this mechanism on many exchanges.

Moreover, ADL is purely related to the crypto market - this is not something you would see in traditional financial markets (if I'm wrong, please correct me). My goal today is to briefly explain what ADL is, what problems it has, and why Bybit (possibly not just this exchange) has done many questionable things with the ADL mechanism, and not everything is clear.

Let me first explain what ADL is.

Actually, there is currently no clear, comprehensive whitepaper explaining how ADL works, lacking clear examples, possible solutions, and plans for the mechanism's development.

A friend recently provided me with ADL data for MANEKI on the Bybit futures platform. I'll slightly modify some numbers (without changing the range and percentage variations) to protect his anonymity.

Automatic deleveraging is a mechanism that closes the most profitable positions (positions opened using maximum leverage) when the exchange cannot continue liquidating other customers' positions without incurring losses.

Let me explain more clearly, in case the previous explanation wasn't clear enough. Suppose there's a massive market fluctuation with a significant price drop, causing many customers' accounts to be forcibly closed. At this point, market liquidity deteriorates, and market impact becomes increasingly significant.

Let's assume we're at a specific moment when only one customer is forcibly liquidated. Suppose this customer's position on $MANEKI is worth 50,000 dollars. For such a position, to avoid forced liquidation, your account needs to maintain 2,500 dollars (because the maintenance margin ratio for this position is 5%).

If your "account value" is below 2,500 dollars, your position will be forcibly closed - that is, the maintenance margin is the cash you need to keep in your account to hold an open position.

Suppose the market continues to fall, and you lose a lot of money on your open position, causing your account to drop below 2,500 dollars - this triggers the automatic closure process.

Let's define some terms:

Liquidation Price → The price of $MANEKI when your account is forcibly closed

Bankruptcy Price → The price of $MANEKI when your account value becomes 0 (if you have a long position, this price should be strictly lower than the liquidation price, because at the liquidation price, your account should still have approximately the maintenance margin balance)

Execution Price → The market order price associated with account liquidation.

Your account is closed through a market order (sell), and the exchange gets the weighted average transaction price for that order.

Here's a very important point - if the execution price (EP) (assuming you're in a long position) is higher than the bankruptcy price (BP), theoretically your account value should not be 0 (because the account value is only 0 when liquidation occurs at BP). However - you won't receive this additional funds - this money is taken by the exchange.

If the exchange can execute the order at a price better than BP, it will collect this surplus (and - according to the documentation - possibly invest it in the insurance fund).

We'll return to this issue later, but now let's focus on a more complex scenario:

Suppose market volatility prevents the exchange from executing an order at a price better than BP. In this case, the account's "account value" will be below 0, and the exchange will use the insurance fund to cover the losses.

The insurance fund is a "reserve pool that the system can use to protect traders from excessive losses in derivatives trading."

Therefore, any forced liquidation order below the bankruptcy price will cause additional losses, and the insurance fund will be used to cover these losses - this fund is typically composed of the exchange's funds or additional profits collected during liquidation.

This is the ADL mechanism we ultimately want to discuss - if the insurance fund cannot cover the losses (because it's empty), the exchange will initiate the Automatic Deleveraging mechanism to bridge this gap.

I want to define it as accurately and clearly as possible: ADL allows the exchange to hedge liquidation losses against profitable customers' positions.

So, if a customer's position is forcibly closed, and the insurance fund cannot absorb the losses, the exchange will find a profitable customer and forcibly close part of this profitable position to compensate for the losses.

Suppose the exchange needs to liquidate a position worth 30,000 dollars: the insurance fund cannot cover the losses, and the exchange ranks customers according to the formula: Profit × Leverage. Assuming the top-ranked customer holds a position worth 100,000 dollars, the exchange automatically closes 30,000 dollars of this profitable position.

No liquidation order will be sent to the market - this won't affect the order book, and the entire impact will be silently absorbed by profitable customers, with no warning.

It sounds a bit crazy - but this is basically how it works.

You could say this is a "over-profit protection" mechanism created by the exchange.

I guess you should understand this mechanism now (I really hope so - I always try to explain slightly complex things in a simple way). Now we can start discussing some questions and issues I have with this mechanism.

What are the strange aspects of ADL?

First, the problem is that we know very little about the ADL mechanism. For example, we don't know when the exchange decides it can't properly execute liquidation orders and must resort to the ADL mechanism. This could happen at any time, without any explanation.

For example, Carol Alexander has a great article about the insurance fund and some volatility events.

There's a very strange point: on May 19, 2021 (all serious crypto traders in 2021 should know this date - if you don't, I suggest making some tea and spending at least an hour looking at the data that day), there was a massive sell-off and forced liquidation event.

But when we look at Binance Futures' insurance fund data, we find that the insurance fund actually grew that day - it wasn't depleted!

I know it should have been emptied because I personally was ADLed on many contracts - clearly, something was wrong with the mechanism. I strongly recommend you read this article.

Let me return to the decision between ADL and market sell orders. Regarding market sell orders and potential slippage: during the massive ADL event on $MANEKI (I assume it was quite large, as I know many people were affected in the past few days), the turnover cost for a 50,000 dollar order (i.e., the cost of simultaneous buy and sell market orders) didn't significantly increase compared to before the massive forced liquidation event.

Recall: for a 50,000 dollar order, the maintenance margin ratio is 5% - even during massive volatility, the turnover cost (not just the one-sided cost) would be lower than this.

For larger positions (and thus potentially greater market impact), you would anyway have a higher maintenance margin ratio.

Next, the crazy part is the difference between the mark price (assuming this is the price at a certain moment in the market) and the ADL price.

In my friend's case, his ADL price was around 0.0033, but the mark price at the time of ADL (i.e., the actual market price) was 0.002595 - a very significant difference.

Even crazier - the 0.0033 price had already disappeared from the market for over 30 minutes when the ADL occurred.