Liquidations 1.2 Parameter Analysis

Hey guys, the analysis / logic behind picking the params for Liq 1.2 is fairly difficult. I think the best we’re going to be able to do on short notice is guess some ranges, and then suggest a conservative option. Please review the below analysis and give feedback, and help brainstorm ideas for a more formal research strategy. This is a very gut-feel analysis so far. Here’s some context for how we’re thinking about it.

Goal: What is a reasonable amount of Dai that can be liquidated without running into zero bids, without accruing significant bad debt to the protocol, giving Vault owners a fair shake recovering their collateral, etc. The mandate here could use some more specificity.

  • The current Dai supply is roughly 450 million. Roughly 300 mil of that was generated from farming, which leaves 150 mil “organic.” This is roughly a 50% increase since 3-4 months ago, which is fairly consistent with intuition. The 300 mil is more or less levered up off of ETH, and generally speaking can’t be relied upon to bid on liquidations.
  • Previously, we used to estimate ~20% of the Dai supply as being freely available for keepers to use in auctions. Of the remaining 150mil, this would imply ~30m keeper liquidity.
  • USDC-A and USDC-B also now offer decent lines of credit. With sufficient time (is 6 hours enough?), keepers should hopefully be able to squeeze out another 20 million Dai liquidity. Additionally, keepers can probably source Dai from Curve or Compound these days, but let’s not depend on that.
  • Analysis based on liquidation price points is ineffective because the distribution changes so frequently.
  • Analyzing keeper liquidity (the optimal research strategy) is just really difficult. One idea would be to check all previous addresses that have bid on auctions in the past and check their on-chain stablecoin holdings. This could provide a rough estimate, but would take a little bit of time, and I want to get some feedback before we go down a research path (maybe there are easier solutions?)
  • Complications:
    • How many keepers currently have their capital locked up in farms?
    • How much of the Dai supply that we think is available for auctions is actually going to unwind Vaults?
    • What is the latest on keeper infrastructure / keeper presence?
  • Parameter ranges: IMO, intuitively, there is no way we liquidate more than 75 million Dai at once. For me that is an extreme upper bound. Lower end is probably 25 million. Midpoint: 30 million from current circulating supply, plus 20 million from USDC Vaults = 50 million. Therefore, a conservative number IMO could be 25-35 million. 40-50m seems “normal.”
  • What’s the downside of having it be too low?
    • Price continues to tank and collateral is auctioned off at worse prices than otherwise would be, leading to minor MKR dilution.
    • This could lead to several PR issues. Users complain that collateral wasn’t sold off fast enough. Or one person got liquidated, their friend got lucky and didn’t. Complaints start filling in to the forums
    • In general, erring on the side of caution (worse liquidation prices) seems much preferable to zero bids. If this is a stopgap solution before Liq. 2.0, then hopefully should be fine.
  • Network congestion concerns
    • Assume a box of 30mil, and a dunk of 50,000. That’s 600 auctions. Let’s ~double that since not every auction will be the full size of 50,000. And each auction takes 3 txs minimum (kick, tend, dent). If my math is right, we’re talking a tx every 5-6 seconds over a 6 hour time span. This seems a bit risky. Of course a keeper can bid on many auctions in the same block, but just want to say this is trickier than it seems at first glance.
    • Given Ethereum is 15 tx/s, this equates to roughly 1-2% of Ethereum network capacity.
  • Suggested parameters for launch: box: 30 million, dunk: 50,000
5 Likes

Just confirming, the spell will be deployed with: box : 30 million Dai, dunk : 50,000 Dai

1 Like

This does have me worried :grimacing:

Hi guys, as I mentioned in chat I would really like to see the box parameter lowered to 15M. I do not think the Keeper ecosystem is robust enough, or the Farmers well enough incentivized or technologically advanced, to muster 30M in liquidity within a 6 hour period.

I know there’s some disagreement on this, so I would like to know if there’s something I could do to help convince everyone? For instance, I’d be happy to reach out to the various entities that I know of running Keepers (or if any of you are in this forum, please chime in) and ask them to pseudonymously submit the number they would be comfortable with.

1 Like

As you pointed out, the best way to estimate keepers reserves is to simply ask them, but then we must hope they are providing honest numbers. If they reached out being concerned about potentially too high box parameter, this tells me there could be some real concern. I would definitely be more cautious if they claimed it is too low. I am not saying keepers are bad players, but definitely they have different incentives, so we must be aware of that. It would be somehow in their interest to lobby for a higher box number so there is less competition and lower bids.

We should also be aware that there might be other keepers with capital ready that we may not be aware of them. One of the ideas we had was also to perform on-chain analysis of labeled keeper’s addresses and see how much stablecoin capital they hold. It is not ideal solution but the only one I could think of, apart from asking them.

Finally, I was thinking, in case of keepers reserves being completely utilized and we start seeing low or zero bids in that 6 hour time span, wouldn’t whale stablecoin farmers just simply want to bid through some sort of UI, such as this one from DeFi Saver? I am just trying to understand how big of an issue is not having enough keepers and if this could be mitigated in practice.

1 Like

Thanks for the reply, Primoz. Even with a manual UI, liquidity and timing could be a big issue. Imagine a scenario where the entire box is filled and 30M Dai needs to be raised in a six hour period. From my estimates, our current Keeper ecosystem would likely be able to immediately provide 15M and would struggle to source the remainder. Even with a UI, are we confident that there are ecosystem participants with 15M Dai that will materialize in such a short time frame?

Since this is only a temporary issue, as liquidations 2.0 should mitigate most capital constraints, I think the safest bet is to make the box lower until the new system is deployed. Playing out “worst case scenarios” in my head, I’d rather have 6 hours of additional price risk on an un-liquidated Vault than several millions of dollars bidding under market.

Again, I’m only one voice here, and as you pointed out it’s really impossible to know what the Keepers actually have on their balance sheet. I really just want to make sure that this parameter is carefully considered and discussed by the community.

1 Like

Tough question, I guess it depends on a) ability to communicate through DeFi quick enough that there is ETH on sale at Maker b) there is still enough DAI liquidity available with low slippage for those who don’t hold it but could buy it with other stablecoins in order to bid and c) these “potential keepers” could utilize USDC-A or B Vaults to mint DAI and bid on those auctions.

1 Like

Yeah, it’s just more unknowns than I am personally comfortable with. But at the end of the day I don’t have the ability to change this… Is there anything I can do to try and escalate this? I am concerned that by time it can go through the normal governance process we will have liquidations 2.0.

Create a signalling request to update the box parameter. Follow the guidelines here. Use some of the previous signals as a template.

If you’re unable to propose something like this due to regulatory concerns, you can either try to convince a member of the community to update the box parameter - possibly one of the keepers you mentioned - or you can convince Cyrus + Primoz that it represents a significant or unnecessary risk and get one of them to create the signal.

If you think it should be done as an urgent or emergency measure this post goes into the details of how to make changes on an expedited timescale.

1 Like

Thinking about the worst part of incident response in another Black Thursday event some of the keeper bidding strategies are to delay bids until the last few minutes of the auction. This means it will be hard to know how healthy auctions are until we see a batch or two finish, and only then can we trigger the liquidation circuit breaker (after taking on more litter in the box).

The good news is, today our exposure is limited with a 30mm box and however long it takes for governance to trip the circuit breaker. Additionally, this is our exposure in an oracle attack too. :confetti_ball:

However, I share everyone’s concerns. We want the perfect box. Too large and we have the problems above, too small and we possibly risk under collateralized Vaults by the time they make it to liquidation. However, in this scenario, auctions will be as competitive as the market would allow, and Vault holders have more time to re-collateralize their Vaults.

Our risks might be better if we set box too small than if we set it too large. I’m leaning towards @g_dip suggestion of a 15mm box. Without knowing the perfect number for box :crystal_ball: , what are people’s thoughts on (risks of box too high) > (risk of box too low)? Is there a counter argument to this line of thinking? If people agree, we may want to lean towards the better set of risks and lower box to a number that is within the ballpark of what we think keepers liquidity is.

3 Likes

Looking at only 6.2 million Dai available in Uniswap V2, I question whether our system could handle much more than 5 million dollars of liquidations in a short period of time. I’m in favor of a smaller box of around 5 or 10 million.

?? This analysis.

What DAI pair on Uniswap are you looking at? How much slippage are you targeting? You also have to look at DAI liquidity on Curve too.

People can also source DAI liquidity from USDC vaults.

Though, people are talking about capital constraints in this thread too.

What do you mean by bidding under market?

That’s a lot of additional price risk. 6h+ OSM, once the box gets full, people can max out their vaults, and they won’t get liquidated. Can choose whether they want to save their Vault later.

It also means, that the auctions that are getting auctioned later, may have a smaller safety buffer for liquidation. And a smaller safety buffer has a much higher chance for auctions to lose money.


I think the main issue being discussing is capital constraints. I think public UIs and awareness makes a big difference. There’s definitely lots of capital floating around and people would be up for buying ETH at a discount.

3 Likes