Outdated algorithm caused $650M excess losses on Hyperliquid, report

Source: CryptoNewsNet Original Title: Outdated algorithm caused $650M excess losses on Hyperliquid, report Original Link: Two months on from October 10’s crypto market meltdown, which saw $19 billion of positions liquidated, Gauntlet CEO Tarun Chitra argues that common autodeleveraging (ADL) mechanisms led to massive losses on Hyperliquid.

In a lengthy post to X, Chitra says an excess of $650 million was autodeleveraged from profitable traders’ positions. The amount, he claims, was 28x more than the potential bad debt facing the exchanges who used ADL.

This “massacre of the innocent” could allegedly be avoided with new ADL algorithms, described in an accompanying 95-page report.

Autodeleveraging on autopilot

Chitra describes ADL as a “last resort” which applies a “haircut” to profitable traders to “cover the bad debt of insolvent positions.”

The 10-year-old “Queue” algorithm is widely used by perpetual futures platforms such as certain head exchanges, Hyperliquid, and Lighter.

However, under extreme market conditions, when ADL is activated repeatedly, “the greedy Queue strategy completely fails.”

The strategy assigns “haircuts” as a function of profits and leverage which, Chitra says, concentrates losses on the biggest winners, while overshooting the necessary amount to be liquidated.

He suggests a “risk-aware pro-rata” algorithm which assigns ADL based on the leverage of each position.

The post recognizes that “a perfect [ADL] strategy does not exist.” However, optimizing for three elements of a so-called ADL Trillema (solvency, fairness and revenue), and running on October 10 Hyperliquid data, the new approach appears to significantly outperform Queue.

Chitra ends by urging for further innovation in the design of algorithmic clearing: “ADL was invented as a band-aid in 2015. We haven’t even begun to explore the design space!”

Hyperliquid’s Response

In response to Chitra’s post, Hyperliquid’s Jeff Yan quipped, “Those who can, do. Those who can’t, fud.”

However, rather than responding directly to the claims of inefficient autodeleveraging, he takes issue with the description of the relationship between ADL and Hyperliquid’s HLP insurance fund.

He accused Chitra of “spreading lies masked by fancy ML terms to sound smart.”

Other Hyperliquid supporters pitched in, pointing to apparent inaccuracies and bias due to investments in competitors.

In the wake of the October 10 crash, Yan argued that “ADLs net made users hundreds of millions of dollars by closing profitable short positions at favorable prices.”

He highlighted that the platform’s ADL queue incorporates “both leverage used and unrealized pnl,” while thanking users for feedback. He also alluded to research “on whether there can be substantial improvements that merit more complexity.”