Maker VaR estimates presentation

Here is the presentation from today’s risk call going over VaR numbers of the maker portfolio.

Here is a pdf version of the presentation
Here is the google slides link if anyone is interested in looking at my notes.

I will provide a youtube link in the comments once that gets uploaded.

14 Likes

Thank you Andy. And thank you for not reading the Risk Disclaimer–in the past folks like Cyrus would have read the entire thing out-loud :wink:

Some alarming stats you dropped on us–but, best to be realistic. Hopefully we will maintain a bullish sentiment via a short-term window to prepare properly, and take the proper measures. Liquidations 2.0 can’t come fast enough…

1 Like

Here is a link to the presentation if anyone missed the call.

Feel free to reach out with any questions I’ll do my best to answer.

5 Likes

What do you recommend as a VaR threshold for the Surplus Buffer? And when you say seek insurance against any VaR breaches—are you referring to institutional insurance, such as Lloyd’s of London, or Nexus Mutual who did not get much traction from the Maker Community? Lloyd’s would be an out of sight move!

Also, can you expand on the idea of higher liquidation ratio Vaults? Business growth vs. Risk management

Idk that i really want to comment on this too much. I think this is a question for governance as to how risk adverse everyone wants to be. I think the proper protocol here would be to start some sort of signal request and then go from there.

I am mostly just suggesting this idea generally. I wouldn’t really be qualified to speak to who might be an appropriate underwriter. I do think that the portfolio is large enough that some institutional insurance agency might be interested in the DAO as a client, but I wouldn’t be able to say how best to set that up.

So there are kind of two questions here. But I’ll post a graphic from the last presentation to kind of explain the thinking here.

The thinking behind high LR vaults would be to try and increase the area that would be shaded in green in the picture above. Basically it means that you would have more of a buffer before MKR holders start to incur losses. The question is what is the appetite for this sort of loan structure. IDK that there is a good answer to that, but I suppose that there might be some appetite so long as the SF was low enough.

But more generally on the topic of “business growth vs risk management”; you may look at those numbers and say this nominal risk is:

So this will leave you with a dilemma when it comes to topics like increasing DCs. Should you raise the DC and continue to raise the nominal VaR or do you decide you need to wait until your cash reserves are bolstered? Do you re-prioritize potential collateral applications to in order to only add collateral types that better diversify the portfolio? Do you consider charging for untapped revenue streams in order to more quickly bolster cash reserves? Lots of things to consider and they often come with trade-offs.

4 Likes

Thank you for proving your views Andy–yes, there are a lot of questions that the community must answer. Personally I am okay with high LR vaults.

With regards to Insurance/underwriter, perhaps @g_dip and @mrabino1 might have an opinion/idea if someone like Lloyd’s of London would insured a DAO, or how to approach them. Not much of an insurance expert myself. It would be reassuring to have someone like Lloyd’s protect the funds depositors place in vaults, and Maker. But then again, it’s a centralized risk? not sure. They even insured Jennifer Lopez’s bum haha

Interesting enough–I came across this article this morning about a startup getting into the insurance business–stating that “The policy is a crime contract covering electronic theft and phishing risk, with cover provided by MakerDAO, an existing investor in the Nayms venture.”

Well I don’t want to see this brushed off so casually so i guess I will provide a little color.

Imagine that i gave you a standard 6 sided die, and i said if you roll a 6 you’ll need to deal with another black thursday type event, but this time it will be at least 10x worse.

That is basically what that VaR80 number says.

Edit: @ElProgreso I didn’t mean to call you out here I read this as “Im ok with high VaR values” for some reason

2 Likes

Is this an annualized risk?

Maybe depends on what you mean by that.

It is the risk that the model predicts some event of this magnitude happening in the next 365 days.

you really just need a custom cds that is acquired in favor of entity that has the dao as the recipient…

2 Likes

Do we locate the issuer of the cds via ISDA? For comfort. Or not necessary.

BTW, this week ISDA launched a digital MSA.

Hi @Andy_McCall thanks for your presentation.

Are this Var values measuring “expected loss of portfolio value” (e.g., ETH-A or ETH-B vaults) or “expected loss for MakerDAO due to unsatisfactory liquidations (value going down too quickly) etc”?

The two numbers are clearly correlated but, I feel, ultimately the Surplus value should depend on the latter value. Because MKR holders are only afraid of liquidations failing, not ETH losing value.

Can you comment on that?

thanks

It isn’t expected loss. The risk team uses that to determine the SFs for a given vault type.

The VaR numbers correspond to the worst losses seen in the simulation. The “tail losses.”

Sorry if this was already discussed, but do your slippage estimates change with Dutch Auctions and flash loans (in the proposed liquidations 2.0 module)?

I think right now the conservative box parameter gives us a good amount of protection – is this a sentient you agree with?

1 Like

:thinking: so the answer here is obviously i don’t really know, but I’ll give you my take just know that this is mostly based upon intuition and speculation.

This might be true. It prevents us from dumping a ton of collateral on a distressed market which might cause further price displacement. On the other hand though it leaves you exposed in a sense as the price could continue to fall as you try and clear the auctions. There might be further work needed in that area. Likely there is some optimal setting here that we might be able to find. I’d be really curious to know how the box might work with auctions 2.0.

So again hard to say. I know @Primoz is really bullish on it and I think it does help in certain respects. So understand that the following is just based on my own views here.

I think it will likely lead to less upward pressure on dai during the auctions, but the slippage model is currently ignoring that anyway. I think it also likely leads to more competition on these auctions, so it could affect how we think about modeling “keeper profit expectations.”

That said, the current slippage model is really trying to model two things keeper profit + how dumping x dollars worth of collateral on the market will affect the price. I don’t really see how auctions 2.0 really has any effect on the later portion of the slippage. Assuming that keepers are still trying to immediately liquidate the purchased position for a profit they still have the opportunity to create short term downward pressure even with auctions 2.0.

So my guess is that it likely helps so long as you don’t need to liquidate more than a few 10s of million worth of collateral, but after that you still are in a situation where the crypto market itself is too illiquid to actually purchase that amount of collateral at a reasonable price.

But again that is just my take.

3 Likes