Morpho and Euler both subscribe to the curator model of risk management. Recently we have seen curated vaults take on excessive risk in the pursuit of outsized returns, only to be met with high standard deviation outcomes resulting in bad debt and loss of user funds.

Similarly, @IMFCrypto has historically experienced liquidation spirals resulting in bad debt, in part due to a misaligned risk framework. Since their unfortunate liquidation spirals, IMF engaged with me to help them develop a more robust risk framework for their vaults. In helping them do so, I have made observations that are inherently prohibitive to risk management, and in some cases introduce or exacerbate risks.

Generally for lending markets, capital efficiency and competition means balancing two preeminent risks; Bad debt (liquidators cannot or will not execute liquidations), and liquidity (lenders cannot withdraw funds for unacceptable durations). @chaoslabs research identifies a third risk; yield volatility risk. I classify this as a performance risk rather than a safety risk, so I won’t dig in to it here.

YQ wrote an article about the Stream Finance saga and provided some technical recommendations for curators and vault service providers. It is a great article, however I see these recommendations as operational more than technical. In expanding on that article and exploring more broadly through my own experience as a curator, I would like to offer some risk-oriented design recommendations, leaning further into the technical space.

A Note on Nomenclature

Deposit Caps - Bad Debt Risk Consequence

The first observation is that @Morpho markets do not allow curators to set a deposit cap. Arguably the most fundamental risk management parameter in a lending market, curators cannot limit the exposure of assets. In developing a risk management framework for IMF and assessing onchain liquidity conditions to recommend allocations, it became evident that capital was (and still is) bypassing IMF vaults and supplying USDS directly to the market contracts, likely because those markets were experiencing high interest rates.

Consequently, some markets became extremely overexposed, the most recent example being the REKT/USDS market. IMF withdrew their entire allocation weeks before the liquidation spiral and IMF-USDS depositors avoided bad debt in the recent cascades, but the fundamental issue is that as risk managers, we have no ability to cap a market’s supply and mitigate liquidation outcomes.

If there is demand, I will write more about the new @IMFCrypto risk framework and how we are rigidly handling liquidation risks to provide the safest USDS yield going forward.

On a side note, it is crazy to me that this isn’t talked about more. Maybe I am in the wrong circles, but I would have thought that curators would be screaming for deposit caps.

Interest Rates - Liquidity Risk Likelihood

The @Morhpo IRM design also inhibits risk management because although it is designed to autonomously manage liquidity risks, the control theory applied is crude and volatile. Control theory can sound complicated so pardon this particular section if it feels convoluted. The IRM is designed to maintain optimal utilization, yet experiences an exponential growth/decay of rates just 1% either side.

This leads to highly sensitive and erratic rates, making it impossible for users to maintain consistently performant positions (perhaps feeding the yield volatility concern). The IRM response is also designed to be normalized around optimal utilization, meaning that integrator compensation ignores the absolute offset from optimal utilization. If a market’s target utilization is 90%, the interest rate will double in 10 days if utilization hits 95% (half way between optimal and max, or 5% above optimal).

In order to return rates, utilization must then sit at 45% for 10 days (half way between optimal and zero) instead of 85% (5% under). The amount of capital that needs to move to achieve this is extreme. In order to achieve a desirable response, control theory suggests smoothing the rate function around optimal and referencing the absolute offset from optimal (rates should half in 10 days if the utilization sits at 85%, or 5% below optimal). Understanding the implications this has on integrator gains over extended periods of time is crucial, but the system will be much more performant.

Oracle selection - Bad Debt Risk Likelihood and Consequence

The next issue is that of oracle selection and redemption rates. It doesn’t necessarily apply to @IMFCrypto, but it is worth mentioning in light of recent events. G from Ethena recently wrote a thread about the differences between oracles pricing from liquidity versus from collateral/ backing. While the piece is enlightening, it needs to be considered in context. When risk curators use redemption rate oracles, risk is transferred from borrowers to lenders. Borrowers get a better experience, having their risk of liquidation due to temporary price dislocations reduced, but lenders take the risk of not liquidating collateral impairment.

This was evident in Stream’s xUSD catastrophe in which some markets continued pricing xUSD at the redemption rate while redemptions were frozen and the price decayed. Curators need to recognize these trade-offs when making decisions. The baseline condition set should protect lenders, always. From a vault curator’s perspective, the price of the collateral asset matters more than the redemption rate, and if the redemption rate is to be used, supply must be capped to correspond with reasonable redemption limits, the same way that supply must be capped to correspond with liquidity depth if using price.

Generally speaking, curators should consider using price oracles to ensure liquidation efficiency and lender protection, and reinforce borrower protection by reducing LLTV where assets are correlated and have peg deviation risks. In some cases, a redemption oracle makes sense, and perhaps implementing a breaker when price deviates significantly may pause markets and allow the situation to be investigated. Reducing the likelihood and consequence of bad debt risk must be the priority, regardless of any yield implications.

Conclusion

There are plenty of great features implemented by both Morpho and Euler Finance that enable risk management, but I think it is too dismissive of Morpho and Euler to say that they are entirely not responsible for recent events. If you want the curator model to succeed, you need to develop the appropriate architecture that actually allows for risks to be legitimately mitigated.

To the curators, you need to identify the gaps in your risk frameworks, and work with the vault service providers to build these mitigations in. Make your risk frameworks as transparent as possible so that users can understand why you deserve your performance fees and why your vault is earning 1% less than your competitors but won’t blow up in the next crisis.

There will be more high-sigma events in the future, and each one offers opportunities from which to learn and develop better products and frameworks.

Optimize Your Designs

Stop building on assumptions. Start building on mathematically proven stability. If you are developing a complex DeFi protocol and require quantitative analysis to justify designs and mitigate risks, we invite you to connect with our team.