Vault (near) zero Dai Bid Liquidation Compensation idea

Assume each vault self-liquidated with a flash loan, plus 13% liquidation fee and auction profit of 3%.

Then all we need to know is the price of Dai.

Dai was expensive at the time, and would have been more so if all vault sourced Dai to self-liquidate (price was already spiking over $1.10 without it).

Let’s say everyone pays $1.20 for Dai.


Vault with 100 DAI debt, holding 1 ETH. ETH price gets to $150, self-liquidate (100 DAI + 13 DAI penalty)*1.03, so owes 116.39 DAI which cost $1.20 each, so $139.668.

Vault gets $10.332 worth of ETH (06888 ETH).

$10.332 back from ETH when it hit $150, so it’s 6.888%.

This is likely higher than in an auction because the price of ETH may have continued falling below $150 during the auction.

The problem is narrowed down to: What would the price of Dai have been?


Two things.

First is that no matter how debt is extinguished (auction, self liquidation) it all contribute to DAI burn. So DAI price was as high it was going to go (1.1ish).

Secondly using the tab change against collateral change is a valid, alternative and useful way to determine compensation.

I worked on another compensation model purely based on the vault Tab (borrowed + interest) and using the difference in collateral before and after the event. In your example I would take the 113DAI tab going to zero and the collateral difference 1ETH-0ETH (0% collateral return). If governance decided that say 10% of the collateral should have came back would simply apply [ 0/1 - .1 ] x (113-0)=-11.3DAI as compensation (only if this number is negative would the owner be eligible to compensation - in this case 11.3DAI). This doesn’t involve any calculation of ETH or DAI price because the collateral liquidated would be based on the value of the outstanding debt.

BTW: The [ 0/1 - .1 ] is basically the collateral deficit.
collateral deficit = ETHafter/ETHbefore - nominal collateral return (decided by governance)

The downside of this is that is pays less than trying to compensate in collateral simply because we are using a percentage related to collateral against the DAI Tab and the ETH price is basically 2.4x what is was when the vaults were liquidated. Example above using 240USD/ETH would basically give that vault the 11.3/240 ETH .04708ETH or 4.7% collateral return.

What is interesting about this model is that it pretty much covers everyone using the same model which satisfies fairness and equitable distribution of compensation. Since the above uses the nominal collateral return against the DAI tab in principle for the same governance price tag (in DAI) we would have a higher nominal collateral return % in the Tab compensation model vs. a collateral based compensation model. In effect the Tab model has the potential to cover more vaults but compensating each vault ‘less’ than they would have gotten in the collateral based model.

As rune said usually something that is a good idea is arrived at by different people. :wink: So you are confirming my ‘Tab’ based compensation plan is probably a good one. What governance will go for. We will see.


First is that no matter how debt is extinguished (auction, self liquidation) it all contribute to DAI burn. So DAI price was as high it was going to go (1.1ish).

Good observation that ultimately the same amount of Dai gets extinguished*.

I see three distinct ways of extinguishing Dai that would each affect the market price of Dai differently (assuming Dai needs to be acquired):

  1. self liquidation – instant market buys of Dai

  2. normally functioning flip auction – market buys of Dai spread of over auction duration (likely to be short since bid time was 10 minutes), and needing twice as much if there are at least 2 auction participants

  3. zero bid flip auctions – market buys of Dai to participate in Flop auctions (after System Surplus is depleted (*), auction delay period elapsed, and auction spread out over days, and again needing twice as much if there are at least 2 auction participants)

In my estimation, self liquidation (or normally functioning flip auctions) would have significantly exacerbated the short-term Dai price spike vs. option 3 (which is mainly what happened) because they would have happened over a very compressed period of time.


I had to think about this a bit and it mostly came out of my liquidation analysis. Didn’t quite matter which way it goes (1-3) no matter how you liquidate the Debt - the debt effectively liquidated (and hence affects to market supply) ends up the same. Now to MKR holders they would have rather pocketed the 13% liquidation fee vs. giving away collateral at a surplus loss that had to be made up with flop auctions.

Now one can look at timing. In the loosest sense the 0-bid auctions would have actually pulled more DAI from the markets driving the price up because the timing of the MKR flop auctions to fill the 5.3M gap was much later than the market event which would have required more DAI to be sourced.

At some point once I get a clean vault history dataset for the MCD history I should be able to determine how much DAI was minted (I already saw (when trying to pull vault data from oasis by hand) a few vaults that minted and sold DAI into the price spike) vs. how much was liquidated via flip/self liquidation etc.

Now if the MKR flops kicked in while markets were still trying to clear flips - that would have make this worse. MKR would not have auctioned because the price offered would probably have been sub 200DAI/MKR auction price cutoff. If you think of the event this way 1-3 literally are identical since they all would be kicking off various auctions to raise the DAI at the same time.