This is in response to @PatrickHmmmmm on discord. Patrick asks a few questions.
Hey Wayne, id love to hear more about the idea before I build the account health system
Would like to build a wishlist for people seeking health notifications. Currently scope is to let people opt into email and text warnings when they hit health thresholds of 60%+
And/or when prices reach a trigger level
That sounds like a really great start and trust me I understand scope of work 
These are just some ideas that i have that may build on what you have proposed in your grant proposal high level concept. I have not seen the full grant proposal.
I’m also thinking about building some interactive tools for people to see what drops in asset prices will do to their borrow position
What do you want to see on the more systemic side
Let’s say for the moment that we are only talking about market risk that is the risk posed to any account or the system due to the movement of prices. In that case, trad-fi has a deep and rich history that should really be used. That is not to say that they have always been right and in fact they have often been wrong however they learned some very valuable lessons that we should remember.
Here are some “well known” observations from tradfi that are pretty much replicated in crypto in my experience.
- When prices are dropping the entire market goes “risk off” and correlations in price movements approach 100%
- Empirical scenario analysis is critical to capture these events
- The assumption of stable distribution of asset returns is absurd
With this context let us consider what that means for compound accounts and the protocol itself.
Accounts should have the ability to adjust prices and see what that does to their portfolio. This is fully independent scenario analysis. It would be better to use the empirical copula to figure out what a change in price in asset X results in assets Y and Z and … This could be called risk factor scenario analysis. It is a much bigger ask with lots of in-the-weeds practical questions. Another method is called “parametric analysis” where a statistical distribution is attempted to be fit to the asset returns. This is mostly flawed on its face in crypto due to the assumption of “independent and identically distributed asset returns”. There are more complex parametric models that may be cooked up but there are so much empirical data (observations) in crypto that parameterizing is not entirely necessary or even good.
The protocol itself has its own health which is represented as the aggregate of all account health within the protocol. This requires some more detailed writing and thought but on its face is quite obvious. There are likely many in-the-weeds issues that will come up when thinking through the math of how accounts aggregate to risk to the protocol. The most obvious question to ask is under what scenario(s) does the entire protocol become insolvent? The insidious part of this question is that the situation is likely much worse than it appears due to resonant frequency liquidations (liquidations causing more liquidations) and leverage throughout the crypto ecosystem due to high leverage perps available at many exchanges now. In that sense there is likely some significantly higher price where there is a point of no return and further liquidations are inevitable.
There are some assumptions built in here especially around the ability and willingness of a liquidator to liquidate in the presence of parabolic gas costs for example (among many others). Having something that isolates price risk is extremely useful in and of itself however.
In summary I think there are a few things worth looking at here
some easy ones
- account health
- collateralization ratio of the entire protocol
- some measure of “health” of the protocol
some harder ones that can and should be applied at the account level and protocol level
- independent price movement scenario analysis
- conditional risk factor price movement scenario analysis
If I had more time, I would have written less (sorry guys)