HLP markets - liquidations after short auctions
A few weeks ago the HLP market on Vega Protocol experienced significant price volatility, resulting in some closeouts and loss socialisation. Although an issue with the price monitoring trigger setup for the market (discussed in the incident report here) meant that the market went into a shorter protective auction than the market creator had intended, the project team have analysed the transactions executed on that market during the window and do not believe any deliberate manipulation occurred, nor that any protocol bugs were hit.
However, one edge case was observed which is worth highlighting. When the incident occurred, the market was initially trading at around 5.5 USDT, with volatility causing a spike up to around 15 USDT in a single jump. This resulted in significant losses for holders of short positions and significant gains for those who were long. However, as these losses for the short holders were often larger than their balances, many of these holders were unable to fully cover their debts even after being fully closed out. This is to be expected with such a large jump, and is what you would expect to happen if this price move reflected the true market price.
However, this then results in the holders of long positions receiving (much) less in gains than they should have, as the shortfall caused by the underfunded short position holders must be socialised between market participants. (In a longer running market the insurance pool would have covered this, however as a newly running market this pool had not had time to build up significantly).
As an unfortunate consequence of this loss socialisation, some keys trading on the market did not receive enough in funds to cover the increased maintenance margin for their positions at a price of 15 USDT vs the margin needed at 5.5 USDT. This resulted in them also being closed out and their margin transferred to the insurance pool.
This behaviour is as designed, though it is obviously not ideal. In general, the risk model and available leverage along with market protections like price monitoring should mean that a well-configured market does not experience conditions like this. Indeed, we have not seen it before, and perhaps would not have seen it had the price monitoring auction lasted as long as intended. We will investigate whether there’s a safe way to handle participants experiencing this edge case better in a future version of the protocol without opening up an avenue for abuse and exploitation.
It’s also worth noting that the margin held by the keys that were liquidated while long in this way (i.e. when they should have been receiving gains) was transferred to the insurance pool as is the case for all liquidated funds.
It would therefore be possible to transfer some of the amount lost back to the affected keys, see the table below, from the market’s insurance pool (which currently contains ~1708 USDT), which may undo the “no fault” confiscation of margin that occurred at least partially, if not entirely, depending on the available insurance pool vs. the total lost. This can only be done via on-chain governance, meaning that someone would need to post a transfer proposal on-chain for the community to vote on.
This post serves as a starting point to discuss if the community would like to do that. We can help with data analysis/APIs/etc. if more information is needed for specific keys.
