Risk Parameter Updates 2021-12-23 and Risk Dashboard Bug Fix

Simple Summary

A proposal to adjust one (1) parameter for one (1) Compound asset.

Abstract

This proposal is a batch update of risk parameters to align with the Moderate risk level chosen by the Compound community. These parameter updates are the sixth of Gauntlet’s regular parameter recommendations as part of Dynamic Risk Parameters.

Motivation

This set of parameter updates seeks to level set assets to a Moderate risk level of the protocol while making risk trade-offs between specific assets. Note that some are different from the original risk level consensus check as market conditions have changed.

Specification

As shown on our dashboard, these changes will slightly increase Value at Risk, but will also increase borrow usage and long-term reserves.

Parameter Current Value Recommended Value
ETH Collateral Factor 80% 82.5%

The increase in reserves as a result of these increases in CFs is $470K.

Results are calculated using the below user behavior assumptions in response to the CF updates. We would note that our platorm is ingesting actual on-chain elasticity and comparing it to these assumptions.

Dashboard

Gauntlet has launched the Compound Risk Dashboard. The community should use the Dashboard to better understand the updated parameter suggestions and general market risk in Compound.

As shown on our dashboard, these changes will increase Value at Risk by $59.3M (10.3%) and increase the borrow usage by 5 basis points. Note that all the VaR is projected to be liquidations that can be safely absorbed by the market.

In addition, we wanted to notify the community that we just resolved a bug with our dashboard data aggregation that attributed the liquidated amount to the borrowed assets instead of to the collateral assets. This caused the dashboard to show a different Value at Risk (VaR) distribution than intended, since the VaR calculation factors both liquidation volume and insolvencies into the calculation.

This does not affect the parameter recommendations we have delivered, and it does not affect our parameter recommendation methodology.

This bug only affected dashboard metrics and not model inputs that drive recommendations. We wanted to give this heads up that the VaR distribution on the dashboard will be different now, but the overall VaR level remains unchanged.

Next Steps

While Gauntlet expects to initiate a governance proposal for this set of parameter recommendations on Wednesday 12/29, we will cancel the vote should changes in our daily simulations dictate it necessary.

While i appreciate enthusiasm, but seriously doubt Compound protocol need to adjust for every couple percent change in some simulation model, especially as often as in 1 month timeframe. Predictability is much more important feature for protocol, aiming at mostly institutional usage. What if in one month model shows that is should decrease in 2 percent? We going to adjust it again?

I’d say rather leave it alone or at least target for quarterly revisit of Collateral factor. That thing ideally shouldn’t be touched often at all, as people rely on stability of Compound when creating their investment models, incorporating Compound protocol.

1 Like

While I like the idea of increasing the ETH collateral factor (CF), I think it needs to be clearly defined what circumstances will result in a recommendation to lower the ETH CF. Unlink BAT, the ETH market has many users, so any change affects a much larger market. Decreasing the ETH collateral factor should be seen as a significant change to Compound.

It would be helpful to unpack the significance of the $633m VAR shown on the dashboard before making any significant changes. If I am interpreting the dashboard correctly, the blue ring is ETH and currently is $540m, and with the proposed change is $590m. That is a $50m increase for $470k more (best case) revenue. If there is truly $540m currently at risk in the ETH market we should be moving to decrease that risk.

Proposal 71 increased the ETH CF to 80%. The forum thread associated with that change showed the VAR collectively after the change was projected to be $132m.

The VAR stat was introduced in the 11/17 parameter changes at $111m with a projection of $132m. The subsequent parameter change on 11/30 had VAR at $76m with a projection of $78m. The following recommendation on 12/11 put VAR at $1291m and a projection of $1385m. My assumption is the 12/11 numbers were affected by the frontend bug you mentioned and I regret not noticing the error (I did not vote on the proposal). The current post’s $574m and projected $633m are a significant increase from 11/17’s and 11/30’s VAR with the ETH market appearing as the primary contributor.

The significance of VAR in the process of managing risk is unclear. Perhaps the dashboard needs to be updated with the statistics that are driving these recommendations or VAR needs to be more clearly defined and scoped.

-Getty

1 Like

Thanks for your feedback, @Sirokko . As a part of Gauntlet’s Dynamic Risk Parameters engagement with Compound, we tune risk parameters so that Compound can grow safely. We would note that a 2.5% increase for ETH is rather substantial given its TVL. Our agent-based simulation platform optimizes the trade-off between capital efficiency and risk in the Compound Protocol - as market conditions change we need to make this tradeoff to ensure that Compound gets higher capital efficiency in safer times, and lower risk in riskier times. User experience is an important consideration in our analysis, which is why we take an incremental approach to updating parameters such as Collateral Factor, to avoid market shocks to the ecosystem.

Thanks @GFXlabs for your feedback and questions. On the topic of lowering Collateral Factors, user experience is an important consideration in our analysis and we would lower CFs when market risks outweigh the capital efficiency gains. In the process of managing risk for the protocol, this means that CFs can increase and decrease depending on market conditions. Our software platform also takes into account factors such as any liquidations that would be caused as a result of CF decreases. On the UI side, we would welcome the opportunity to collaborate with the Compound developer team to make it easier to notify users of CF changes.

On the topic of VaR - to clarify, the frontend bug had no impact on the overall protocol VaR numbers. The VaR increase you mentioned was driven by increased market volatility during that time period (market downturn in early December) and lower collateralization ratios amongst the top borrowers. We note that all of the increase in VaR is projected to be liquidations that can be safely absorbed by the market. We value your feedback on updating the Dashboard - we are working on breaking out the VaR between liquidations and insolvencies so that users can see the split, as this is something that users have asked for in our user studies.

If historical volatility and average collateralizations ratios are the drivers behind CF decisions, it would be helpful to see those displayed on Gauntlet’s site.

If all of the VaR can be “safely absorbed by the market,” is the value at risk? I would think “VaR” would be the amount of funds that would not be safely absorbed by the market and turn into underwater positions.

I think this question went unanswered. What are the key stats that are driving your recommendations? VaR increasing due to “increased market volatility” & “lower collateralization ratios amongst the top borrowers,” but your recommendation is to increase the CF?

-Getty

Thanks @GFXlabs - we are in the process of making developments to the Dashboard to help inform users of our methodology. Regarding VaR - we are making updates on the VaR calculation and will push those changes as well as publish a deep dive on VaR early in this coming year.