Small Vaults Liquidations & Incentives

The main goal of this forum post is to introduce and explain the issue at hand with small vaults not being able to get liquidated and provide some suggestions for actionable next steps that the Maker Community can consider.

Introduction

On November 8, the B.Protocol team observed that small, undercollateralized Vaults are not being liquidated as they should in the Maker Protocol. Further, the cost of splitting a vault into small vaults is small compared to the liquidation penalty at current gas prices. Vault holders may start using this tactic to avoid having their position liquidated. The preliminary evidence suggests that Keepers are choosing not to liquidate these Vaults (as opposed to not detecting them), since larger Vaults have been consistently liquidated in recent months.

It’s important to note that this has been a known possibility that has been discussed several times in the past:

Considerations & Concerns

Based on recent observations, it has become apparent that small (debt close to the dust limit), undercollateralized Vaults are not being liquidated by Keepers. Further indications suggest that this is due to Keepers not seeing these liquidation opportunities as profitable:

  • The existing Keeper infrastructure can detect split Vaults as well as normally created ones.
  • The lack of liquidation was observed for at least one Vault created normally (not via splitting).

The operation to split one’s Vault into small, un-liquidatable Vaults is currently economically sensible (about 1% of a Vault’s debt, much less than the liquidation penalty).

  • Vault holders may start taking advantage of this to avoid liquidation—with governance needing at a minimum three days to raise the dust limit, there is plenty of time for a Vault holder to wait for the price to increase before un-splitting their Vault.
    • Someone could release a tool that allows all Vault holders to take advantage of this operation.
  • Liquidations 2.0 includes an incentive that governance can adjust to directly reward liquidators; currently, the dust parameter is the only accessible tool to address this issue, at least at the protocol level.

Technical Explanation

  • Technical explanation of the dust parameter and its impact on the system
    • dust: the debt floor for a specific collateral type.
    • A new Vault cannot be created with less debt than the dust specified for its collateral type.
    • A Vault that has non-zero debt lower than its dust limit cannot be modified except in a way that either raises its debt above dust or reduces its debt to zero.
    • The situation described in the previous bullet point will arise when the dust value is increased for a collateral type with Vaults already opened for less debt than the new dust value.
  • Technical explanation of the LIQ2.0 Keeper Liquidation Incentives system
    • Liquidations 2.0 includes incentives for Keepers that liquidate Vaults in the form of DAI rewards paid directly from the Protocol to the liquidator.
    • The technical summary of Liquidations 2.0 explains in detail how these rewards are calculated.
    • This feature was added after feedback from the community highlighted the lack of incentives for liquidation as a major concern in the existing system.
    • The list of preliminary liquidation parameters for control of liquidation rewards can be found in the following forum post: https://forum.makerdao.com/t/liquidations-2-0-preliminary-risk-parameters/4850

Next Steps

In the short-term, the MakerDAO community may consider increasing the dust parameter to combat this issue. Further, there has been some activity pushing forward this short-term solution. Last week, @Primoz (Risk Domain Team) proposed a Signal Request to increase the dust value. The Signal Request has since closed with no clear winner. Based on the Signal Request results, three options had almost equal votes (500 DAI, 250 DAI, 100 DAI (no change to current dust value)). Primoz stated that the next steps to determine whether the Maker DAO community wants to increase the dust parameter is a ranked-choice on-chain Governance Poll with the same options listed in the Signal Request. The Governance Poll will be published on Monday, November 16.

As for the longer-term plans (Liquidations 2.0), a solution for Keeper Liquidation Incentives has already been included as an explicit reward to incentivize liquidations.

11 Likes

This is Yaron from B.Protocol.

Agreeing with @charlesstlouis that this is a known issue, and in principle it is just a matter of setting the appropriate dust parameter.

Our point of view is that analytically figuring out that number is quite hard. And more generally the current DeFi liquidation incentives are hard to quantify, which is a concern both for the liquidators and the platforms.

However, at this post we would like to comment more from the position of a platform that offers a Vault management interface, with over 100 Vaults and $5m in ETH.

The dust value is a very painful UX issue. A rational user would like to experiment with a new platform (like B.Protocol, but also if he is using MakerDAO for the first time) with small amounts, before he gains more confidence.

Hence, in the absence of a clear mathematical solution, we would like to raise the possibility of leaving the dust value as is, and applying some practical techniques until the 2.0 upgrade. E.g., the foundation could daily (and even hourly?) test keepers for 100 dai liquidations.

On the protocol level, it is of course our belief that all current DeFi liquidation systems would benefit from a professional backstop with clearer incentives. And encourage MakerDAO Vault owners to manage their Vaults via our interface (and by doing so, share some of the backstop proceeds).

6 Likes

I am sympathetic to this point. Can we entertain more drastic changes to the system that would address this? For example, the collateralization ratio could be set as a function of the size of the vault. For example, something like effective CR = CR’ * dust / MIN(dust, vault value) where CR’ is the nominal CR and the effective CR is the one that is actually applied to the vault.

1 Like

A vault with debt $120 currently have a collateral of around $180. Do you think that if the collateral would be, say, $300, it will improve the incentive of the keepers? Currently the maximum theoretical profit is $180, which is very decent. The (imo) is that the expected profit is smaller (and imo also unknown). Making the collateral bigger would improve the expected profit? (as opposed to the maximum profit).

1 Like

Fwiw my personal opinion is that the long-term solution here really is backstops like B.Protocol or KeeperDAO that allow liquidators to play a cooperative game that preserves some margin for themselves. Competitive auction models IMO squeeze Keeper margins too much to create the kind of robust liquidator ecosystem DeFi protocols (not just Maker) need.

5 Likes

Moreover, every incentive that you can give directly to keepers, as opposed to miners, and any additional certainty, will eventually benefit also with the users. In the form of smaller liquidation penalty, and/or smaller collateral ratio.

At the end of the day MakerDAO ecosystem is very robust, and have MKR as the ultimate backstop, but for DeFi to scale beyond the typical 130-150% collateral ratio, our believe is that more cooperative approach is needed, and the B.Protocol approach, in its non-permissionless/decenteralized version, is already quite standard in CeFi, and allow it to offer x100 leverage (101% collateral ratio).

1 Like

Plus 1 on this one! Maybe link, or provide a pop-up message that tells small Vault users to “check out” B.Protocol–we’ll be doing them a really big favuor, IMO. Maybe, we can even get a referral fee :stuck_out_tongue_closed_eyes: J/K. But ya, B.Protocol is super user friendly.

6 Likes