[Signal Request] Should we change box & ttl auction parameters?

Hey all,

Referring to concerns brought by @g_dip regarding the box parameter used in Liquidations 1.2 and his discussion with some of the keepers in the ecosystem, they advised to decrease the box value from 30m to 15m. Lower number means that auctions clear more efficiently (lower probability of low bids), but we risk additional price volatility for potentially postponed auctions due to market conditions. However vaults having postponed auctions could unwind in the meantime, which means that setting box conservatively can still lead to certain positive outcomes.

On the other hand if we are concerned about the box being too high currently (keepers don’t have enough capital) we would normally hope for people from the DeFi ecosystem to bid on auctions using for instance UI from DeFi Saver.

Another idea that was raised in chat rooms was to mitigate potentially lower box with lower ttl (reducing auction bid duration from 6 hours to 3 hours). This means that 30m auctions, despite having a box of 15m, would still be cleared within 6 hours and would hence reduce price volatility exposure. However 3 hours might not be long enough for all keepers to react, relative to geographic location of their operations.

Pros for lower box:

  • More participants per auction, higher bids
  • Vaults still have time to unwind before liquidated

Cons for lower box:

  • Price volatility exposure for non-liquidated vaults that get liquidated in next batch
  • Even if are overestimating keepers’ reserves currently, we might be underestimating liquidity that could come to auctions from other users using DeFi Saver liquidation service

Pros for lower ttl combined with lower box:

  • More balanced effect of increased price exposure concerns and faster clearing of auctions

Cons for lower ttl combined with lower box:

  • 3 hour bid auction duration might not be enough for all keepers to react
  • Decrease box to 15m and leave ttl at 6 hours
  • Decrease box to 15m and decrease ttl to 3 hours
  • Keep box at 30m and ttl at 6 hours.
  • Abstain

0 voters

This poll will close on Friday, September 11th and if there is a majority it will go on-chain on Monday, September 14th.

2 Likes

Just a quick note to anyone who is confused that this isn’t following the usual timeline. I spoke to Primoz and we feel that we should try to get this on-chain on the 14th due to the recent market movement, but we’d like to leave more than a few days for polling. So this is closing on Friday.

Thank you for putting this up, @Primoz. I would like to give some clarity on why I am voting for “Decrease box to 15m and leave ttl at 6 hours.”

(1) 30m (current setting) likely exceeds the current Keeper ecosystem’s available liquidity. Over the weekend we saw two WBTC vaults with a cumulative 50m Dai in debt come within 5% of liquidation. Should his have happened, 30m Dai would have been required to bid on the first set of auctions. This means that within 7 hours (1hr OSM delay + 6hr auction ttl), the Keepers would need to source this capital in either Dai or USDC (to use the USDC-A and B facilities). Based on my conversations with participants in this group, they would not have these funds readily available in this timeline. A lack of Dai liquidity can potentially result in 0 bids if more Keepers are not able to quickly join the ecosystem or the existing Keepers able to source emergency credit lines. Either way, this is not an enviable position. Therefore I am voting to decrease box to 15m, which I believe to be within the current Keeper ecosystem’s available liquidity. I should note that this would be a temporary change, as liquidations 2.0 should remove the need for this kind of cap. As Primoz pointed out, the downside to this decision is that bad debt can sit in the system for a longer period of time.

(2) I believe the current ttl of 6 hours should remain the same. The downside to cutting it in half, is that we risk a technical issue (or even sleep schedules) resulting in 0 bids. For example, if Keeper X represents 30% of the liquidity, and they are knocked offline for whatever reason, that extra 3 hours could be the difference between the collateral clearing at the correct price or 30% lower than it should have. I do not think that pushing the potential clearing time of bad debt by 1 hour (one OSM update) is such a large risk that we should reduce the Keeper’s time to source funds and bid - I believe the latter to be the greater risk. If the community believes that the reduction in box needs a reduction in ttl, I think 5 hours (which mitigates the additional one hour) rather than 3.

I posted some of my concerns in the other thread regarding lowering the box to 15m with a ttl of 6 hours.

Open source UIs also allows anyone to be a keeper. There is a ton of DAI liquidity out there, as you can see by all the yield farming operations. Currently, there’s 80m DAI providing DAI/ETH liquidity to farm Sushi. With public awareness, we can get people to bid on Maker auctions to buy collateral at a discount.

A 15m Box doesn’t seem sufficient considering our current DAI debt size.

There is a lot of additional risk when the box gets filled. People will be able to max out vaults and not get liquidated. They 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.

The current issue is between capital constraints, and the increased risk with a lower Box size.

A 15m Box size doesn’t seem sufficient, it adds more risk. Capital constraints risks may be overblown.

I agree with everything you said about the risks of setting the box too low, but I disagree with this one last statement. I think keeper capital risk mixed with having the box set too large puts us in a similar situation as Black Thursday. Specifically, I think the risks you articulated from the box being set too low are a better set of risks than if the box is set too high.

In the former scenario, the price drops below the protocols ability to return capital to the Vault holder and the protocol takes on bad debt; however, when the liquidation happens we get a competitive bid on the collateral. The bonus here is that Vault holders get a chance to save their Vaults, and we are more resilient against the more likely occurrence of a flash crash and recovery, as opposed to a permanent collateral crash. Meanwhile, the protocol takes on bad debt, but that bad debt value isn’t the entire position as in the case of 1 wei bids. This could happen if a collateral instantly goes to 0, but I can imagine no auction scenario that mitigates this risk.

In the latter scenario, keepers cannot muster the capital in the six hour window and we see bids far below market prices, possibly 1 wei bids again. This means the Vault holder looses everything, and the protocol takes on all that bad debt for a large recapitalization with MKR. This is the Black Thursday scenario, and it’s a far more likely risk with disastrous consequences.

I’m interested in any counter-arguments to my conjecture that risk of box too low < risk of box too high.

1 Like

Do you have an opinion on the shorter vs longer ttl?

There are so many factors for ttl:

  • keeper risk of holding a bid for ttl (impacts magnitude of early bids).
  • how quickly can a particular collateral have its capital recycled (WBTC).
  • network congestion for bids on ethereum (this is a big one).
  • number of transactions needed for a given dunk size and overall Vault profile.

Having not balanced all of these risks personally, I would say 6 hours ttl is a good default. I would not start fiddling with ttl without doing a thorough risk assessment on the risks above and any others we can think of.

P.S. To give everyone an idea of network congestion on Black Thursday. I was paying 2x+ gas price, and it could only execute 0.7 transactions a minute. So applying this to auctions for a single address, that’s (.7 tx a minute) * (60 minutes in an hour) * (6 hours) = 252 transactions for a single address. For auctions on BT, I think there were 4000+ auctions with each taking minimally kick(), tend(), dent(), and deal() so the keeper can recycle capital. For competitive auctions, that’s a lot more transactions. Even if the solution is to go wide with many addresses or web3 users, how much does the gas block size rate-limit liquidations? How much now that DeFi is exploding with many other protocols where users are competing for that space? I’ll be working on LIQ-2.0 if you need me.

4 Likes

I agree with this and it seems to be the sentiment I hear from the Keepers as well. I am not firmly opposed to changing it from 6 hours to a lower number, but I don’t think we should do it in a bundle with a change to box. Better off changing these parameters in isolation with their own analysis.

Poll closed, winning vote was to decrease box to 15m and keep ttl unchanged at 6 hours. Parameter changes will go into on-chain poll on Monday 14th September.

2 Likes

Hey Primoz, just so I understand the process, this needs to be polled now and then put into an executive? What is the timeline for actual implementation? The protocol is taking on a lot of risk by having this parameter where it is, in my opinion.

2 Likes

The vote goes into on-chain poll on Monday and into executive vote on Friday next week. Based on posts here I didn’t get a clear message that changing this parameter is an emergency thing and if you look at the text under the poll, the timeline was described from the beginning and nobody had any objections.

1 Like

No ill will meant from my end! I missed the timeline in the original post, sorry about that.

1 Like

Hey @Primoz … Since this seems relatively uncontroversial and it looks like there may be an executive tomorrow anyway, would you consider expediting its inclusion to tomorrow? After speaking with more Keepers last week, I believe the liquidity situation is worse than most think and that 15M is actually an aggressive, not conservative number.

Expedited Executive Vote: 2020-09-14 Included in the expedited executive vote on Monday.

1 Like