Kujira Liquidation Limitations

We currently have over 3.3B worth of bLUNA deposited on Anchor and only around $26M in the liquidation pool. This shouldn’t be a problem under normal circumstances, but we are again opening ourselves up to a situation where liquidations stop during a flash crash as a result of limiting the premium to 30%. There have been a lot of fancy theories as to why liquidations stopped during the May crash for a short time until TFL had to step in. It’s really this simple. If the premiums were free floating and set at market rates, there will always be enough liquidators around to keep Anchor solvent in even the most extreme situations. In small dips, low-mid single percentage premiums are probably enough to entice liquidators. In crashes harder and faster than the one we saw in May, 30% clearly wasn’t enough and maybe 50% might not have been either.

If we are going to let the market decide the premium, we should not artificially limit how small or large the premiums are. We will again be forced to rely on TFL to clean up the mess of a flash crash without enough liquidators in the marketplace under the current setup. Let’s create a free market for liquidations before it becomes a problem again. Let the premiums range from 0%-100%.

1 Like

Great thinking in free-market terms.

I tend to agree but seem to think we need a balance of both free-market and protecting protocol solvency. Setting liquations % discount to what would be below the 100% collateral would risk protocol losses. It’s kind of like a lockdown limit circuit breaker that is needed for markets to cool.

1 Like

I have to agree with bitn8, while in practice it seems like a good move then 30% is not a random number, it’s to help maintain solvency for the protocol, during a sustained downturn of the market, making it available at higher premiums would probably help in the long term but during a flash crash it could open up the protocol to heavy losses than would be avoided by simply not liquidating.

The main reason for allowing for premiums to exceed 30% is to protect protocol solvency.

The situation that requires premiums of greater than 30% is the flash crash. Think about what happened back in May. Liquidations were occurring in an orderly fashion until the LUNA-bLUNA spread became wider than the premiums. At this point, liquidators stopped liquidating. If Anchor’s liquidation bot did NOT step in, Anchor was at risk of becoming insolvent and going under in May. It is an existential threat as a result of the premiums being limited.

In May, Harpoon was running a bot that performed the liquidation and swap from bLUNA->LUNA in a single transaction. They had a minor oversight that caused the LUNA->bLUNA arb insanely profitably momentarily as the LUNA-bLUNA spread widened far beyond the bLUNA liquidation premium a few times during the crash.

If Anchor is to ever be truly decentralized, removing artificial limitations on liquidation premiums is a necessity. Without it, Anchor either remains under a central authority or becomes insolvent and dies during a flash crash.

TLDR; the Anchor community has 3 long-term choices:

  1. Remove liquidation premium limits and allow for true decentralization/reliance on TFL.
  2. Keep liquidation premium limits and require a TFL liquidation bot to continue running forever to ensure Anchor stays solvent.
  3. TFL steps away from liquidations with the current liquidation parameters in place and a flash crash makes Anchor insolvent.

As more liquidity is added, especially with White Whale’s impending launch this spread is going to tighten significantly and I foresee these spreads being much tighter and certainly less threat. Also having more protocols plugging into the liquidator queue can also help create more demand for liquidations.

That said, I still think you raise a few valid points. The community should look deeper at the 30% limitation and debate where it should be but there isn’t that much room to move it without allowing for protocol losses. No liquation limit also has protocol loss/solvency issues if liquation premiums take liquidation into negative TLV area where premiums spike into 40+% territory. So if we did set a range, the question starts to become how much do we balance in a premium that can create protocol losses, especially since the deeper issue here is flash crash liquidity, that circuit break-action would typically resolve while markets stabilize and liquidity comes back in.

I think the current moves to get more liquidity and access to the liquidation queue from protocols and average users is going to drive liquidation liquidity and partly mitigate this issue. I think the focus should be more on this than the actual percent.

I’m not sure what you mean by this. TVL cannot go negative with if liquidations are occurring in an orderly fashion. The total borrowing on Anchor CAN exceed the collateral deposited in a system in which liquidation premiums are restricted. It that were to occur, Anchor becomes insolvent, depositers cannot be paid out in full, ANC goes to 0, Anchor is effectively dead.

Limiting liquidation premiums is extremely risky. If we do limit, we need to look at something higher than 30%. We’ve seen 30% fail spectacularly in May. People on the Harpoon UI stopped liquidating when the bLUNA->LUNA spread was not enough to overcome the discount on bLUNA via liquidations. Harpoon’s bot that was frontrunning the Harpoon UI users. The Harpoon bot made a few mistakes and liquidated some large wallets and took a close to 50% haircut on some transactions when going from bLUNA->LUNA, which isn’t enough to cover the gains from liquidations. Harpoon/Kujira is more sophisticated now and we can’t rely on their mistakes to perform liquidations at a loss to keep Anchor solvent.

The amount of money deposited into Kujira for liquidations is less than 1% of the collateral deposited. During the flash crash era, liquidators had significantly more than that available for liquidations and it wasn’t even close to enough to keeping Anchor afloat without TFL’s bot stepping in. With all due respect, White Whale’s impact on the LUNA-bLUNA spread is unlikely to be significant. We have bigger players in the ecosystem who will beat them to that arb. The need for Astroport and a stableswap invariant AMM formula for LUNA-bLUNA cannot be overstated.

And by the way, how many people are aware of Harpoon’s team running a bot to frontrun Harpoon UI users’ liquidations? Is it a little funny that they were complaining that bots were making it hard for the Harpoon UI users to perform liquidations when they themselves were running one of the 3 bots that were performing the vast majority of the liquidations? Is it also a bit interesting that they only started this after they received a 22,000 ANC grant that was supposedly used for the purpose of creating a Harpoon website for Terra users? Are we that happy that the same team is running Kujira? Does anyone want to see a competing product run by someone not affiliated with Harpoon/Kujira?

1 Like

Correct. So I think we are fundamentally agreeing here that the issue is liquidity but both propose different ways to solve that. Curious to hear other people’s take on this.

I guess it will be left to see what White Whale does. I’m hopeful for the project but again I admit we won’t know until it is out for a while.

Agree with this 100%. But if the spread is low in liquidity and goes outside the bounds set in invariant curves, the spread gets wider than the current CPMM xy=K curve for the same spread. So again, for invariant to work we need way more liquidity in the pair to create a tight market. Hopefully, WhiteWhale, Astro, and other Dex bring way more liquidity to allow stable invariant curves for this pair. Other ideas for liquidity include, CDP in LUNA-bLUNA when tendermint finally builds out flashloans functionality.

Yes, that’s what I was getting at. I think we need to have more for redundancy and hopefully also find a way of incentivizing protocols to build in liquations.

1 Like

Maybe it’s just me, but I was thinking anyone except Harpoon/Kujira as an alternative. I don’t believe there’s any major problems with the current setup unlike the original crap that was the Harpoon UI for users that had no hope of beating bots.

Do you think it would bother people to know that the guys running Kujira/Harpoon were the same guys running a liquidation bot that frontrun Harpoon users and caused users to attempt hopeless liquidations and pay gas fees?