Black Thursday Response Thread

Instead of the voted 4% SF proposal - let’s put the proposal of 2% SF ASAP to fix the peg which is 5% off now. I doubt 4% will be enough and we will lose much time.

1 Like

If there is a docker container, I can make a DNP for it so that dappnode/avado users could run keepers.

2 Likes

I don’t know if this was mentioned above yet (will catch up in a moment), but I wanted to voice my thoughts on one issue.

Thinking slightly longer ahead, for Vault users whose remaining collateral was sold off at zero… I know that they take on and accept the risks involved, but this shouldn’t have happened to them.

Imho, I believe it would be a show of good faith from MKR governance to absorb some or all of their losses. They don’t necessarily need to be paid out in a lump sum. One idea - somehow allocating them a portion of the SF as it trickles in.

Not urgent, but I think it will be a good look.

Thoughts?

6 Likes

I think this is a bump in the road, one to learn from, and ultimately will strengthen the system. Dai is still in its infancy. An expensive lesson, but learning this when Dai is ~100M is cheaper than when Dai is 2B.

5 Likes

Definitely something worth discussing, feels like a sensible thing to do, but on the other hand it’s probably going to mean printing more MKR which there may not be the most appetite for.

More of a longer term issue than short too.

1 Like

Governance Plan and Forum Response

Recap

  • The market experienced a severe downturn, ~30% over 24 hours.
  • Gas issues complicated matters by making it difficult for vault owners and other ecosystem actors to interact with the protocol.
  • Like with previous market events, Infura/node issues further complicated people’s abilities to interact with their vaults, keepers and oracles(?)
  • A large exchange halted withdrawals due to gas prices
  • This resulted in individual vaults becoming undercollateralized and having the balance of their Dai debt sent to the “bad debt” pool. This ate through the $500k surplus buffer, before triggering another $4.5million of liquidations.
  • In the rush to pay down dai debt the peg destabilized.
  • Of primary concern are the events surrounding the Keeper ecosystem:
    • Keeper activity, already historically very low, did not provide adequate coverage during collateral auctions.
    • It appears that one or more keepers, in the absence of any competition, placed lowball orders for collateral

Solution

It is the Foundation Risk Teams recommendation that the following parameters be added to the executive vote this Friday.

  • As per the existing winning polls:
    • Sai Stability Fee: 9.5% => 7.5%
    • Dai stability Fee: 8% => 4% (applies to ETH and BAT)
    • Dai Saving Rates Spread = 0
      • Therefore DSR of 8% => 4%
    • GSM for SCD is getting activated
    • Migration contract debt ceiling: 30M => 10M
    • SCD debt ceiling: 30M => 25M
    • MCD Eth debt ceiling: 150M => 110M
    • Global MCD debt ceiling: 125M
    • Additional parameters:
    • Flip auction TTL: 10 minutes => 3 hours
    • Flip auction lot size: 50 ETH => 500 ETH
    • Flip auction maximum duration: 3 days => 1 day
    • Flop auction TTL: 10 minutes => 3 hours

The change in the TTL for the flip and flop auction parameters are designed to give keepers any additional time needed to overcome network issues such as the ones seen this morning. The increase in lot size will also lower the amount of auctions (and, thus, transactions) that will need to take place. A significantly lengthened TTL means that the auctions are likely to reach the maximum duration. A lower maximum duration would help counteract long auctions.

These options will allow the community to reduce the impact of an additional market downturn while it has the opportunity to continue to debate further, more refined solutions, in the coming weeks.

EDIT / ADDENDUM:

All, due to the highly evolving nature of the situation, we’re proposing some changes to the above parameters.

  • The GSM for SCD is no longer part of the spell. It creates too many edge cases unknowns and puts SCD at considerable risk.
  • The Flip and Flop auction TTLs are both being increased to 6 hours. We are aware that there are some volatility issues due to the long lockup, but right now the priority is that it’s difficult to sync up nodes at all. We should consider erring on the side of node safety.
  • There will be a Debt Auction Delay being included for 6.5 days. This is for a variety fo reasons. First, in the current situation we probably shouldn’t complicate our situation with additional debt auction worries. Additionally, this gives any prospective investors time to collect their ducks in a row to make a strong bid for MKR on Wednesday midday (as opposed to the current Saturday morning at 4am for west coast time).

EDIT / ADDENDUM 2:

  • Flip Tau is will be reduced to 6 hours. Having the Tau be higher than the Ttl just adds significant overhead to monitoring the auctions. With ttl = tau, they will be significantly easier to process.

EDIT / ADDENDUM 3:

  • MCD Eth debt ceiling being reduced to 100M instead of 110M in order to target a ~20 M buffer.
9 Likes

In general these all seem like great changes.

A couple of points:

Flip auction lot size: 50 ETH => 500 ETH

Does this price out keepers?

Flop auction TTL: 10 minutes => 3 hours

If we put this up on Friday it probably won’t come into affect before the FLOP auctions start.

Finally, there is fairly heavy support for us putting up an executive today instead of Friday. Has this been considered?

Don’t think this one is a good idea

2 Likes

@cyrus

With the current SAI supply @ < 20MM does it make sense to reduce this number even more since we are trying to sunset the platform? Maybe we should just drop this down further to $20MM

See you soon @LongForWisdom, Take care !

Some notes, there are some issues when you increase the TTL to 3 hours, with auctions potentially lasting 24 hours.

Keepers will require a lot more capital to process liquidations. With a shorter TTL, keepers can keep processing liquidations, recycling capital and offloading risk onto secondary markets. With a longer TTL you will need to be able to access a lot more keeper capital.

Other options, which probably aren’t immediately available.

  • There could be a minimum bid price (like the flop auction), means shorter TTL.
4 Likes

Looks like the proposed changes from Cyrus will make it into an executive tomorrow (although on the call subsequent to this announcement and based on the polls in this thread, we aim to get the executive live tonight.)

2 Likes

I think there should be question about DSR spread, question about DSR and SF are ambigous

1 Like

Yes minting MKR to sell is basically diluting holders (in the case of equities the shareholders). This is one way to raise capital that markets don’t like. They would rather see a bond issue. Which may be a better way to raise DAI capital from the DSR markets than just letting it fly away by lowering the current DSR.

The problem with this is that the deposit withdrawal mechanics would have to be different because you want the DAI deposited there to have a term (say 2 years). Offer a tasty FIXED rate of return (say 6-8%) over 2 years and then pay back the loan out of the SF fees over the next 2 years and/or sell MKR depending.

Actually the more I think about it the DAI DSR-2 deposit contract could literally not allow people to withdrawal until a specific date and the interest would accumulate as per normal DSR operation. I am just not sure how to implement such a beast into the system

BTW: The problem here is DAI pricing risk which depositors may not be hot to deal with. Perhaps USDC deposits could be considered. Either way bond issues (debt) are typically how companies attempt to raise cash vs. dilution but both are valid financial tools used by companies.

YES I have been advocating putting the SCD dc pretty much at the outstanding within a 1-2%. I see absolutely no reason at this point NOT to do this.

This above is one of the best things I have seen. I love the idea people can do this from the web but honestly I think a better way is to pool and split earnings as overall costs on this operation to all parties can be kept low.

I think 85% might be a bit aggressive and I would target 75% honestly but even so I would still supply funds.

I will add as a response to LFW post on executive changes.

I agree with other comments the auction changes 50-500 seems pretty extreme as well as the time increases. What I wonder is if we can get a sizable pool up and running on the normal auctions. Maybe a double of everything 50-100, time from 10min to 30min vs. 500ETH and hours or a day until a working keeper pool can be set up and operational. Then at least we will know the size of our liquidity provider of last resort.

I also would like to see the SCD DC go to $20M vs. 25M Everything else looks perfect.

1 Like

Yes

Is it possible for us to quote the flip/flop auction lot sizes in DAI instead of ETH?

If keeper returns are measured in ETH, they may not be able to compete on the gas market with trades measured in USD during downturns.

2 Likes

@MakerMan got it. Thank you for the valuable feedback, and handling many questions from the community this afternoon on the Maker Community Call! Would love to meet you one day in person!

Once we plug the leaks by a) introducing improvements to the Collateral Auctions b) onboarding more Keepers, lets figure out how MakerDAO could emerge from the crisis stronger than we were just a week ago.

Moving forward, lessons we all learned today are highly important when it comes to considering new collateral applications to MCD.

Specifically, I would like to bring forward four recommendations:

1. Encourage low LTVs for new collaterals during early days of such collaterals in MCD - Even when collateral value is not as volatile.

Where does this recommendation come from? During the last 9 months I had an opportunity to prototype risk model, calculate stability fee, LTV, debt ceiling (as an independent risk team) in pilots of a new collateral candidate (accounts receivables aka invoices). We were successful at originating real-world loans to trucking companies using DAI using their invoices as collateral (privately sourced DAI deployed to trucking companies who pledged off-chain collateral and borrowed DAI from a legal entity, just without MCD). Based on this experience in finding a product/market fit for MCD in traditional capital markets, I believe we should encourage new collaterals not to attempt replacing most commong legacy financial lending products (which often have very high LTVs). Instead, we should encourage new collateral applications (most of which are in the originating business) - to focus on a “low LTV product-market fit” - a niche in the respective lending markets where borrowers are comfortable pledging said collateral and borrow DAI from MCD at MUCH lower LTV ( such as home lines of credit - i.e. 50% to 70%) and not lending products with high LTV (such as mortgage loans at 80% or 95% LTV).

Example #1: Accounts payable collaterals (invoices of businesses). These guys will argue that 90% LTV IS safe because invoice factoring companies offer 98% LTV (they buy invoices outright). We must argue and encourage such applications to find a niche in the market where borrowers are OK to borrow at 50% LTV.

Example #2: Real estate collateral - we must encourage collaterals to find borrowers who want a “line of credit” type of product (who are OK with low LTV product) instead of a “mortgage” type of borrowers (these folks expect 85% or even 95% LTV product).

2. Define what a “Minimum Viable Keepers” market should be before we allow them into MCD. For example, make sure new collaterals have at least 3-5 Keepers specializing in, and native to this collateral (for example Keepers who understand the collateral, or enough “off-chain auction” participants collection agencies specializing in buying invoices,

3. Conservative debt ceiling allocations. Start with small debt ceiling, and only increase debt ceiling with growth of # of Keepers and history of participation in auctions.

4. Custom Auction Parameters for each collateral. Adjust parameters through governance based on risk/size/liquidity. Think custom Collateral Auction parameters for each collateral (for example:
4.1 liquidation penalty of 10% for real estate, 15% for account receivables, 20% for BAT, 25% for RenBTC

4.2 Consider activation of collateral-specific lot size

4.3 Consider activation of collateral-specific delays

4.4 Consider activation of DYNAMIC liquidation penalty. Liquidation penalty does not have to be static. It could be adjusted according to the risk this LTV and this collateral brings to the MCD system. For example, we can tie liquidation penalty to a bonding curve LTV (for example 50% LTV for collateral X = 10% liquidation penalty, 70$ LTV for the same collateral X = 20% liquidation penalty).

I hope that community considers recommendations above with the goal to reduce probability of Collateral Auction collapses in the future.

4 Likes

Yes. Let’s plug the leak first, but we should also make this right for vault holders. Even if it takes a few months to create a solution, we should announce a plan to do so within the next couple of weeks. It’s imperative for the ecosystem that all participants (the DAO, Keepers, and Vault owners) have their incentives aligned.

2 Likes

2. Define what a “Minimum Viable Keepers” market should be before we allow them into MCD. For example, make sure new collaterals have at least 3-5 Keepers specializing in, and native to this collateral (for example Keepers who understand the collateral, or enough “off-chain auction” participants collection agencies specializing in buying invoices,

I suppose it depends upon the collateral type, but for token based items, I’d vote that we have a readily accessible interface for interacting with the auctions. For example, for ETH-DAI, there should be a website I can go to that lists current auctions that I can bid on in my web browser with something like metamask. Otherwise your liquidity hinges on as few as 3 keepers actually being online, running the software, etc.

3 Likes