The recent drama has shown the fragility of this current model and some of the parameters chosen.
In the longer term, there should be work done to change how the protocol assumes risk – engaging Gauntlet, RiskDAO, or Goblin Sax (the DAO that does underwriting on NFTfi) to do modeling, and also some architectural changes.
It’s not clear that changing the HF for liquidations alone is enough to prevent a situation like this from occurring in the future.
In the mean time, there should be work done to show to the community that work is being done to derisk the protocol, shore up confidence, and provide breathing room for the protocol.
Here are some suggestions:
Disabling new lending from the UI
This would be a low lift way to show that action is being taken, and if there’s a need to transition to a new model, helps alleviate this. Given the low number of new loans at this time, this would only have a minor impact.
Increasing the Reserve Factor
Currently the reserve factor is at 65%. This should be increased to 75%, or slightly below where the current reserve factor is (~82%).
The purpose of the kink is to encourage additional borrowers to deposit, yet with a 37% deposit APY at time of writing, the only recent large depositor is Mandala Capital, with many other depositors pulling out.
The high borrow APY of 46% (net of $BEND Rewards) also increases risk of liquidation in addition to the backdrop of declining NFT prices.
Instead, at 75%, the borrow APY would be 20%, decreasing the liquidation risks, while still providing a strong APY for depositors (~15% APY, net token rewards).
Decreasing Bend Rewards for borrowing
Bend Rewards for borrowing were important for bootstrapping the protocol, but given that BendDAO is already the largest platform in this segment, they’re only dilutionary at this point.
This has also hurt the protocol in that they’ve masked the amount of risk that borrowers are taking on from the protocol.
Implementing a locked staking solution
Given the large fluctuations in deposits and withdrawals, there should be a deposit staking program for which in exchange for increased rewards (whether it be via Bend or treasury funds) – deposited funds would be locked for an x duration – this could be as short as a week, or as long as a month.
Improving messaging from BendDAO
This is probably the most important one – this is a crisis of confidence, and BendDAO needs to be clear in what is being done.
Some potential ideas:
that there are OTC buyers lined up which could potentially alleviate concerns of a cascade.
materials that describe how the liquidation process works
dashboards around bad debt
Reducing the liquidation period to 24 hours
Incorporating what @diego mentioned below, derisking liquidators. This could revised further downward if needed, but 24 hrs would be a good step in this direction.
Ultimately, these are just my thoughts, but would be eager to get feedback or additional thoughts on how these changes could be done.
Great suggestions Will – thanks for writing the proposal.
One of my main concerns is that the liquidation mechanism may not function as well as intended under significant stress. A malfunction of this part of the system could pose a major risk of bad debt for the protocol, which may, or may not, lead to a permanent loss.
As explained in the docs, liquidators’s bids have to be at least 95% of the floor. There’s also a 48-hour grace period following a default, which means that the successful bidder, if the borrower doesn’t pay back their debt, would effectively be buying a long put on the NFT with an expiration two days after the default. However, if the borrower wants to repay during the grace period, they would have to pay a 5% fee to the liquidator (this would effectively be the premium under the put-option analogy).
The number of NFTs at risk and the low liquidity of the market, though, may make these incentives for liquidators not strong enough to effectively protect the protocol from bad debt, which has the risk of leaving a permanent gap in the DAO’s balance sheet.
This is why I think the proposal could be extended with changes to the liquidations parameters. Some ideas include:
Reducing the 48-hour grace period to something much smaller, in the order of a few hours.
Increasing the fee paid to the liquidator to something greater than 5% (effectively raising their max upside).
Relaxing the requirement forcing liquidators to bid at a price above 95% of the floor. Potentially, just allow them to bid at a price at (or marginally above) the balance of the debt.
A risk, however, is if the floor drops below the size of the debt, since the protocol requires bids to be at least of the size of the total accumulated debt. In that case, the incentives to bid become progressively weaker as the floor drops further.
Providing $BEND rewards to liquidators (potentially re-routed from those originally for borrowers). This could help make up for the risks that liquidators are taking and has the direct effect of incentivizing bids.
I really believe that the these ideas, as well as William’s, do a good job at alleviating some of the risks currently faced by the protocol. I ask community members to urgently take a look at them and to explore ways of helping the protocol deal with the real risk of (potentially permanent) bad debt.
Here is another good suggestion from user uetani-arigato.eth:
Moving LTV down to 70% needs to happen. 80% is still too high to make sure that assets entering the danger zone for Bend get rapidly liquidated and the DAO made whole.
That said, this doesn’t have to happen all at once, and it shouldn’t happen as a step function.
DAO can vote to set a new LTV and a timeframe under which it will move to that point: For example, 70%, shift starting September 15 and moving linearly from 90% to 70% over the following 100 days.
This gives borrowers some time to get their loans in order before the shift happens. When the shift does happen, it happens at a rate of 0.2%/day.
During this period, pay 100% of the interest received to Depositors or reduce Borrowers’ interest. The amount that BEND would have taken can be split in some way between the two. Yes, veBEND holders get a haircut for 100 days. Their other option is to hope/pray that BEND doesn’t completely collapse.
Update the front end to make it explicitly clear to borrowers during this time what their LTV ratio is and how many days they have before they get liquidated.
I think we will make a proposal to improve the protocol today.
I’m Alex, Co-founder of Cenit Finance (Startup whose main products are simulation engines both for lending and NFT-lending risk analysis).
We have been reading these posts and we have to say that the suggestions proposed go in the right direction in our opinion. Of course, it is hard to set the exact parameters without running the simulations and getting the setups that lead to BendDao to the lowest VaR.
We wanted to give you some data-driven insights, and we have applied a slippage model to the BAYC situation that gives us a rough idea of how many loans would be un-incentivized for liquidators, that means floor < size of debt. The slippage model is based on new listings added to the secondary market.
Here I attach a spreadsheet where we can see 3 potential scenarios with a fall in floor price, -5%,-10% and 15%, and how these floor price falls would trigger different sizes of spiral.
A caveat is that we have been using a generic model instead of a specific fitted for BAYC, and the results might differ somewhat due to this.
We can see how in a scenario where there is a 15% floor price decrease, the liquidators would have no interest in liquidating at least 22 BAYCs.
Hope this helps you, and if you want me to have a look at other collections (potential) spirals, let me know (or any other way we can help).
Thanks for your shaing Alex, we made BIP 10 to let the community to vote to change the NFT liquidation line, and we will make a web page for showing for data on BendDAO.
Disabling new lending from the UI
Sorry we didn’t do that, except for serious bug which will make our users lost money by hacking, stop the deposit and withdraw or any lending function will cause more panic, that’s not we want to do.
Increasing the Reserve Factor
We need more discuss about it, and a new proposal to change it.
Decreasing Bend Rewards for borrowing
Bend Rewards for borrowing were important for bootstrapping the protocol, but given that BendDAO is already the largest platform in this segment, they’re only dilutionary at this point.
This has also hurt the protocol in that they’ve masked the amount of risk that borrowers are taking on from the protocol.
Implementing a locked staking solution
I personal would like to a more freedom protocol without lock, more lock = less freedom, that’s not why I am here for crypto.
Improving messaging from BendDAO
Yes, we have a emailing system, which can provide email notification for health factor warning, and we may try more service like dc @ or APNS to get a better messaging system.
We can discuss more about these improvements here, would like to hear more from you.
Reducing the 48-hour grace period to something much smaller, in the order of a few hours.
Yes, BIP10 is going to make it 24hours this week, and the community can vote it to 12hours or 4 hours.
Increasing the fee paid to the liquidator to something greater than 5% (effectively raising their max upside).
I think max(0.2eth, 5% of the debts) is quite enough for collection like BAYC, if we make it too much, most of the NFT holders will not repay half of the debts to keep it. Maybe improve the 0.2eth would be
better, 5% of BAYC is like 3.5eth this time, quite much.
Relaxing the requirement forcing liquidators to bid at a price above 95% of the floor. Potentially, just allow them to bid at a price at (or marginally above) the balance of the debt.
BIP 10 make it down to 80%, community can vote to 75% or 70%. We are trying to find a solution to remove the limitation of fist_bid > debt, would post in the forum once we finish the draft, you can help to review
A risk, however, is if the floor drops below the size of the debt, since the protocol requires bids to be at least of the size of the total accumulated debt. In that case, the incentives to bid become progressively weaker as the floor drops further.
Yes, let’s keep making the bid liquidity better by removing the limits and providing downpayment of auction etc, and let more liquidators to join the auctions.
Providing $BEND rewards to liquidators (potentially re-routed from those originally for borrowers). This could help make up for the risks that liquidators are taking and has the direct effect of incentivizing bids.
I think this panic has nothing to do with the $BEND tokenomics, the main reason is that we didn’t show how much the NFT debt are in auction and how much the floating bad debts, 200+ eth NFT debt and dozen of floating bad debts will make 15000+ eth panic withdraw. And the 95% limit of bidding price is stupid, we removed it, and make the liquidation bar from 90% to 80% or lower, make the 48hours to 24hours or shorter, and increase the base rate of ETH deposits, I think we make it this time, we can adjust these parameters from time to time in the future by risk model, let’s build risk model with @Alex_Cenit together.
Hey! For some reason I missed this ping, sorry about that.
Sure, we could prepare a first proposal of how our modelling would work and check it out with you for further iterations. If you find that approppiate, just give me the OK and I will get on it
