Do Not Resuscitate — Contingency Planning For Your CDP in a Bear Market

Do Not Resuscitate (DNR) is a legal order to which gives advance notice of a patient’s wishes that he/she not be administered CPR or advanced life support in case their heart were to stop or they were to stop breathing on their own.

Someone who issues a DNR in anticipation of being hospitalized or incapacitated does so because they perceive the physical and emotional cost of heroic life saving measures to be too great. It is a lower bound a person sets on their own quality of life. It is a threshold at which a person’s rational incentives align in favor of total loss. This is a useful framework to consider CDPs in a bear market if you happen to be in a situation like mine.

I don’t mean for this post to be macabre. Please absorb this in the intended spirit, which is to give you an additional mental model and help you achieve better hedging while using a MakerDAO CDP in a downturn.

In my last post I told you the story of how I paid off $50K of my mortgage using a CDP in February of 2018, right as the 2018 bear market started to take shape. When I began the experiment, I was overcollateralized by about 300%. But as the value of ETH started to collapse, my CDP got close to the liquidation price. Unfortunately, my timing for this experiment had been bad.

Furthermore, and almost more importantly, the mental framework for this loan, in my mind, was “mortgage”. The original intent and spirit of the loan completely guided my actions, as if I was still making decisions that could impact the roof over my head, despite the fact that I had extricated myself entirely from the oversight, judgement, and penalties of the retail banking system. I decided to play it super safe and sold some US Treasury bonds, then paid off the entire 50K in one payment. Unwittingly, the bear market forced me into increasing my total home equity, though in the process, I used up the last of my available fiat reserves.

My CDP was now free and clear, however, and ready to borrow from again. I began to daydream about whether or not I might try using the CDP for leverage to borrow more Dai.

By July of 2018 crypto prices seemed to stabilize and I began running through the game theory of CDPs in general. I’d love to unpack my process here, but it was extensive and deserving of a post unto itself, so I’ll jump to my conclusion which was that there was an overwhelming amount of logic that supported taking on the maximum amount of leverage possible, and a paucity of logic against it. So that month I began borrowing Dai again, but this time the mental framework was “gambling”. A very different mindset.

At this point the price of Bitcoin was in the low to mid $6000s. I assumed we were at the bottom and obviously I was way wrong. This however, is where my own experience uncovered a hidden game theoretical aspect of using CDPs. This edge case comes with a few very specific prerequisites, all of which applied to me. But even if your situation doesn’t quite match mine, this dynamic still applies. Therefore I think it’s important for all CDP users to understand how the DNR way of thinking applies to CDPs, and to have a DNR price in mind when interacting with the MakerDAO system.

At the start of my gambling (aka leverage) experiment I had only ETH to gamble with, but no more dollars I was comfortable losing. This should start ringing some bells for some of you, as I think there are a lot of ETH holders out there like me, who can’t put more cash into the system, but we want to put these ETH bags to work!

As recently as last summer, I was of the firm belief that no matter how low the price of ETH went, that it would eventually return to its pre-bear levels of +$1K. Now I’m not as certain, but this underlying belief of a return to higher values was a primary motivator to set a DNR.

By July of 2018, my total ETH portfolio was worth just under $150K. I wanted to borrow the same amount that I’d borrowed for my mortgage (50K), but even putting all of my ETH into a CDP wouldn’t get me to a “safe” 300% collateralization. For security reasons, I also didn’t want to put all of my ETH into a CDP.

MY PLAN

I lowered my loan target to 35K DAI and lowered my personal collateralization comfort zone from 300% to 200%. Then I pulled the trigger.

Mid July CDP snapshot:

Dai debt: 35K.

Eth price: $465

Collateral locked: 150 PETH.

Collateralization: 200%

Liquidation price: $350

Liquidation Cost: $39,550 (35K plus a 13% penalty)

True Collateral: 85.05376 PETH

I define True Collateral as the amount of ETH (PETH) that I’d lose in the event of an instant liquidation, at any point in time. This is a critical distinction because even though MakerDAO requires 150% collateralization of ETH loans, the entire 150% is not liquidated if you fall below that threshold. What is most likely is that your actual debt amount is liquidated, plus a 13% penalty. You can quickly calculate this number by dividing your total DAI debt by the spot price of ETH, then multiplying that number by 1.13. (Btw this number does not include your accrued interest, but interest in the MakerDAO system is negligible especially over shorter time frames).

Plugging in my numbers: (35,000/465)1.13=85.05376

85.05376 ETH, is what I stood to lose when my leverage experiment began, and this represented $39,550 in dollar terms at the moment of that snapshot.

The late summer pullback began towards the end of July, and on July 31st ETH closed at $419.60. I was now 179% collateralized, and in the Red Zone. Instinctively, I started piling more ETH into the CDP to protect it and bring down my liquidation price.

My first addition was 12 ETH.

July 31st snapshot:

Dai Debt: (still) 35K

Eth Price: $419.6

Collateral locked: 162 PETH.

Collateralization: 194%

Liquidation price: $324.07

Liquidation Cost: (still) $39,550

True Collateral: 94.25643 PETH

If you compare these two snapshots, you see that the US dollar cost of a potential liquidation has not changed. But the penalty in ETH terms has increased by 9.20267 ETH. Assume for a moment that we live in a Universe A where my CDP gets liquidated at the first snapshot, then compare it to Universe B in which my CDP gets liquidated immediately after adding 9.20267 more ETH. In both universes, ETH then returns to its former high of $1,417.38.

In both universes, the liquidation costs me the same amount in dollar terms, ($39,550.00) at the moment of liquidation.

In Universe B, however, I have lost more ETH, and therefore $13,043.68.00 in value by the time ETH returns to its former ATH.

What I realized at this point was that if the only CDP-saving tool I had at my disposal was adding more ETH, then eventually I would run out of room. If I pushed my liquidation price all the way down to my personal limit, and then got liquidated, I stood to lose much more value if ETH ever returned to its ATH.

It’s like a game of chicken. The more ETH you put into your CDP, the longer you can keep 100% of your collateral. But every time you push the liquidation price lower by adding PETH, you increase the penalty should you run out of ETH.

Put another way, if you think that the price of ETH is going to inevitably fall below your ability to save it then rise very high afterwards, the best thing you can do is get liquidated early. But no one has perfect knowledge, and you’d rather not ever get liquidated at all, so the DNR price gives you a rational middle ground.

DNR Price — This is a liquidation price floor, below which you will no longer defend your CDP. It is a lower bound at which point the financial incentives are better in the event of early liquidation, and when the specter of a more extreme liquidation, with much more collateral on the line, appears possible.

It’s important to note that setting a DNR price is a gamble, but so too is adding more collateral to your CDP if you have limited collateral. Each move has consequences. Setting a DNR is the more conservative approach, which is probably counterintuitive. It means being emotionally prepared for a large loss of collateral even though you are in command of the resources to temporarily hold onto it.

Ultimately there was no formula I could come up with to calculate a quality DNR. There were too many numbers that were mere guesses, and they’re all relative. One thing you can calculate, however, is what your lowest minimum liquidation price is, if you were to stake all of your ETH. To calculate this, multiply your Dai debt by 1.5 to get your US$ liquidation threshold, then simply divide that number by the size of your total ETH holdings.

I had 318.2795 ETH at the time, so plugging in my numbers we get this equation:

($35,000 x 1.5)/318.2795=$164.95 lowest minimum liquidation price

To put that number in context, I had to think in terms of narrative, and ask myself what ETH price would represent a crack in the prevailing narrative that ETH might be about to recover. What was the price that would make everyone say “if it doesn’t turn around here, then I fear we’re going a lot lower”? It turns out that there was a very handy reference for this exact number. The average liquidation price across all CDP’s in the MakerDAO system had emerged as a quasi-prediction market on the collective thinking as to ETH’s price floor. By August of 2018 the average liquidation price was $289.00. To bring my liquidation price that low would require a total of 181 PETH.

Using $289 as a benchmark, I had to ask myself this question: If we break below $289, could we go even lower than $164.95. If the answer is yes, then it’s time to set a DNR price at $289. Let’s say, alternatively, that I had enough ETH to bring my liquidation price down to $5. In that case I wouldn’t have thought a DNR was necessary. In order for this to be a useful tool, you really have to figure out if you fall into a narrow range where you could conceivably run out of room.

August 11, 2018 snapshot:

Dai Debt: $35K

Eth Price: $318.15

Collateral locked: 181 PETH.

Collateralization: 165%

Liquidation price: $290.05

Liquidation Cost: (still) $39,550

True Collateral: 136.3558 PETH

I remember the first two weeks of August, because that’s when I realized that we were going lower, and that even though I had more ETH to add, it was time to let go. On Aug 13th, between 7 and 8am my CDP was liquidated.

What I find thrilling in how this unfolded is to know that I actually would have lost all my ETH if I hadn’t picked a DNR price. That’s a very black and white indicator of how useful it was to me. But there’s another exciting prize that I discovered on the way down.

Remember that you are drawing Dai, and either holding the Dai or purchasing other assets. Once the price of ETH drops more than 13% below your DNR, then you’ve effectively shorted ETH for a profit. In other words, once the price hits DNR-13%, you’ve successfully borrowed a higher value of Dai (penalty included) than you could get if you simply sold the same amount of ETH at the spot price of DNR-13%. In this way, just borrowing Dai is a very effective hedge against falling asset prices.

Postscript

Since publishing this piece, someone on the Makerchat proscribed another method for saving your CDP. This warrants a mention both for users to keep in mind, and also to explain why I didn’t do it.

CDPs are always overcollateralized, so by definition you already have the means to expunge your debt, only it’s locked away in your CDP. If you use the “free” function, you can unlock some of that collateral and convert it to ETH. Then sell ETH for Dai, and use the proceeds to wipe some debt. This in turn will allow you to free more collateral, which you can sell and then repeat the process. This would be tedious and require frequent repetitions as asset prices continue to fall, but theoretically you could keep it going until your entire debt is paid off (so long as asset prices don’t collapse faster than you can free collateral and sell it).

The downside is that this method creates a taxable event for every sale of collateral. My one rule for this experiment in leverage gambling was that I would not incur any more fiat expenses, including taxes. This type of intervention therefore wasn’t an option for me.