Dynamic Risk Parameters


A proposal for continuous market risk management to optimize yield, capital efficiency, and mitigate depositor losses.


For almost two years now Gauntlet has formally and informally worked for Compound to perform market risk assessments, contribute to treasury management, optimize incentives, calibrate risk parameters, and upgrade the protocol. During that time Gauntlet has been able to refine our core models and agents specifically for autonomous interest rate protocol’s like Compound.

As the protocol continues to decentralize to the community our position is that dynamic risk parameters are a vital component to growth. Most protocol upgrades and maintenance impact market risk of the protocol. For example, the seize function and liquidator behavior. Or the introduction of Chainlink Price Feeds which has and will continue to facilitate the onboarding of new assets. How should the community reason about initial borrow caps? When should collateral factors be raised or lowered? How do individual assets and their parameterizations affect insolvency risk?


In the following sections, we will outline the case and goals for dynamic risk parameters. The initial proposed scope has target metrics Gauntlet aims to improve. Those metrics are:

  • Risk-adjusted yield for Depositors
  • Capital efficiency for Borrowers
  • Mitigate Depositor losses

Gauntlet will improve the metrics above while controlling for protocol insolvency risk.

Illustrated in the governance example below are the benefits from a previous parametrization initiated and executed by Gauntlet. Additionally, we describe two initial areas for optimization that have been identified.

Compound Proposal 039

The ZRX, BAT, and WBTC Parameter Update governance proposal sought to change collateral factors with a primary focus on lowering WBTC. See the full thread for details.

To measure impact of this change we can look at the total WBTC liquidated on both the Aave and Compound protocols during the weeks of April 18-24, 2021 and May 17-23, 2021. The total liquidity available on the Compound and Aave was similar during these weeks. In addition, the volume of WBTC liquidated on Compound was less than 10% of the volume on Aave in April, which was closer to our parameter change. From there we mapped Compound positions from 2021-02-21 against the subsequent price movements. If users held unchanged positions, which is not uncommon, from before the collateral factor update through single-day price drops of 20% in April and 41% drop in May, then there would have been ~$7M to $9M additional collateral available for liquidation on each occasion.

Capital Efficiency for New Assets

Currently, the collateral factors for AAVE, LINK, MKR, SUSHI, and YFI are conservative. As the supply of these assets grows, ensuring any individual asset does not contribute outsized risk to the protocol is key. Existing simulation outputs suggest increasing collateral factors for all five assets by approximately 15% is optimal. Doing so would allow users to borrow an additional $31M in assets.

Making early capital efficiency improvements like this are possible because Gauntlet runs daily off-chain simulations. Informed by market data (liquidity, slippage, etc.) we will adjust collateral factors lower or higher as needed.

Reserve Factor Support

Gauntlet will also support reserve factor parameterization which is a key lever in driving revenue and growth (increasing yields/reducing interest paid). Previous conversation surrounding Reserve Factor Standardization has been had but no further analysis has been performed into the optimal settings to track default probabilities. Gauntlet concurs that:

While a secondary parameter for risk, the reserve factor is a primary parameter for revenue and growth of the protocol. For example, when changing borrow caps consideration should also be given for the new optimal reserve factor.


  • Risk Parameter Updates

    • Coverage of all markets except Legacy (e.g., WBTC) and Deprecated (e.g, SAI, REP)
    • Supported Risk Parameters: Collateral Factor, Close Factor, Borrow Cap, Reserve Factor, and Liquidation Incentive
    • Market conditions will determine the frequency of updates. For that reason, no SLA will be preset.
  • Communications

    • Risk parameter change steps:
      1. Forum post (e.g.,Reduce COMP emissions by 20%)
      2. Community discussion and revision
      3. Off chain polling
      4. On chain vote
      5. Post-mortem
    • Quarterly, Gauntlet will poll the community to determine the preferred risk tolerance of the community. The outcome of this vote will determine the risk and capital efficiency tradeoffs Gauntlet will target.
    • Monthly forum posts and participation on community calls with explanations of risk parameter changes and any anomalies observed including but not limited to:
      • Discord Developer & Twitter Spaces Community Calls
    • Risk Dashboard (refer to the next section)
    • Quarterly Risk Reviews will provide a detailed retrospective on market risk.
  • Out of Scope

    • Protocol development work, (e.g. Solidity changes that improve risk/reward)
    • Formalized mechanism design outside of the supported parameters.
    • In line with keeping the scope small, Gauntlet will not look to manage the following at the outset:
      • Enabling or disabling a currency for borrowing
      • Setting interest rate strategies
      • Optimizing COMP emissions

Risk Dashboard

As part of this engagement, Gauntlet will build a Risk Dashboard and API for the community to provide key insights into risk and capital efficiency.

Please note, all numbers are for illustrative purposes only and do not reflect the current or possible future state of Compound.

The dashboard focuses on both the system-level risk in Compound and the market risk on an individual collateral level. Our goal is to help convey our methodology to the community and provide visibility into why we are making specific parameter recommendations.

The two key metrics are Value at Risk (VaR) and Borrow Usage.

Value at Risk conveys capital at risk due to insolvencies and liquidations when markets are under duress (i.e., Black Thursday). The current VaR in the system breaks down by collateral type. We currently compute VaR (based on a measure of protocol insolvency) at the 95th percentile of our simulation runs assuming peak volatility in the past year. We do this using Compound’s current parameters as well as after modifying the parameters to the Gauntlet Recommendations.

Borrow Usage provides information about how aggressively depositors of collateral borrow against their supply. Defined on a per Asset level as:

where U is the utilization ratio of each user:

We aggregate this to a system level by taking a weighted sum of all the assets used as collateral.

To show Gauntlet’s impact, we measure these using the current system parameters and expected results (based on our simulations) if Compound were to implement the parameter recommendations suggested.


Gauntlet charges a service fee that seeks to be commensurate with the value we add to protocols. Gauntlet also wants to provide a strong signal of our alignment with the protocol. Using our prior COMP Contributor Grants proposal we propose a service fee using the Contributor Comp Speed grant functionality. At the start of every quarter for one year Gauntlet will create a proposal to update the service fee payment in accordance with the forumla below.

The formula to calculate Gauntlet’s service fee has four components:

  1. An asset multiplier to track risk management complexity
  2. A proxy for capital efficiency
  3. A marginal base fee
  4. VWAP (Volume Weighted Average Price) of COMP

The asset multiplier calculation is log(Number of Assets, 10)*. New assets on the protocol add complexity to risk management. While the market risk optimization problem does not grow linearly, consideration should be taken when onboarding assets.

The most straightforward proxy for capital efficiency is the total borrowed for risk-managed assets. Capital efficiency is realized by borrowing demand. The total borrowed amount is calculated as the 30-day average and rounded down to the nearest $1B.

Gauntlet’s risk management marginal base fee is derived from a conservative estimation of the impact from dynamic risk parameters.

Marginal Base Fee Total Borrow
10 bps $0 - $5B
5 bps $6B - $10B
2.5 bps $11B - $15B
1.25 bps $16B - $20B

The VWAP of COMP for the previous 30-days. Whether the price should be fixed or calculated quarterly, different communities have different opinions on how this aligns incentives. We will defer to the preference of the community but will default to calculating quarterly.

*Gauntlet quarterly service fee denominated in COMP (table above calculated at $464)

Growth and drawdown examples

*Log value is the minimum of the tier range except in the “<= 10” column, where it is 10. For example Column “21-25” returns log(21,10)

** When Total Borrow < $3b, there is no basis point fee. The formula is log(Number of Assets,10) * $1,200,000 / 4 )

About Gauntlet

Gauntlet is a simulation platform for market risk management and protocol optimization. Our prior work includes assessments for Compound, MakerDAO, Liquity, and Aave. Gauntlet’s continuous parameter optimization work includes Balancer, SushiSwap, Benqi, Aave, and Acala.

Thanks to @tarun, @wfu, @shaan, @jmo and many others for assistance on this proposal.


@inkymaze thank you for sharing this proposal on the Community Developer call this morning; a standard process to monitor & tune the risk parameters of the protocol is a long time coming (and one that I think the community should take!).

A few preliminary questions:

  1. With respect to the metrics that Gauntlet intends to optimize (risk adjusted yield, capital efficiency for borrowers, and mitigating depositor losses), how will these be tracked? Will they be on the risk dashboard that Gauntlet plans to create, to monitor them over time?
  2. Proposal 49 updated the liquidation mechanics to reduce the “cascade risk” of certain markets; has this functionality been included in the simulation models yet, or will it be prior to beginning dynamic risk parameterization?
  3. How do you envision setting an acceptable “VaR” relative to reserves? How do you plan to include the community in this decision?
  4. How are the proposed fees being set? Have community members weighed in on these?
  5. Does Gauntlet plan to hold or dispose, vote or delegate its proposed COMP payment?

Initially the dashboard will track capital efficiency via Borrow Usage and risk via Value At Risk. From our initial user studies, these were the key metrics users wanted to see.

In addition to the dashboard, we will follow the outlined communications plan to capture anything that isn’t yet represented (i.e. Risk adjusted yield).

There are additional features on our roadmap like a visualization of users with undercollateralized loans and collateral liquidation ‘depth’. We plan on iterating on this dashboard over time and adding new features that we identify through continued community engagement.

Proposal 49 routes a fraction of liquidation incentive from liquidators to the reserves. This change effectively allocates 5.2% of liquidation to liquidators and 2.8% to the reserves, which increases the protocol’s ability to recover from insolvency by growing the backstop liquidity, but reduces the incentive for liquidators. We will update the effective liquidator incentive to 5.2% in the simulation to accommodate the change.

Our current simulation is mainly focused on modeling insolvency risk in one day. Considering the average liquidation size relative to the sizes of reserves, the 2.8% of liquidations added to the reserves in a day will likely not have an immediate impact in such a short time frame. However, tracking the amount of reserves over time and forecasting the growth rate of the reserves due to parameter changes can definitely help community members to better understand the protocol’s liquidity backstop. Forecasting the reserve growth rate is not in our initial scope, but we will evaluate how to support this in Q1 2022.

Our primary goal for simulations is to standardize VaR across all assets, to ensure no subset of assets adds disproportionate risk to the protocol. We will target a similar system-level VaR to the current risk parameters as the moderate risk level recommendation. Additionally, we will provide aggressive/conservative risk level recommendations by targeting x% of capital efficiency increase/decrease.

As our philosophy is avoiding non-quantitative decisions, we don’t decide on what “acceptable” VaR relative to reserves is. We will facilitate the community’s decision-making process by creating quarterly off-chain polling for the community to decide this high-level objective. Community members can check our risk dashboard to get an estimate of VaR relative to reserves to understand the protocol’s ability to recover from insolvency events.

As mentioned above, Gauntlet seeks to charge commensurate to the value provided. This means measuring our work against the target metrics, communication objectives, and deliverables like the Risk Dashboard. We encourage the community to evaluate these items proactively and on a regular cadence ahead of our proposed quarterly fee update.

We have sought consultation from various community members, not solely our investors, and we encourage all others to use this thread to weigh in on the proposed fees.

Gauntlet is a firm believer in Compound’s mission and growth. As such, we plan on holding COMP tokens and self delegating for governance votes. Depending on Gauntlet’s cash flow needs, including but not limited to tax payments and operational expenses, we may need to sell tokens at a future date. For reference, we have not sold any of the COMP initially granted to us in Dec. 2020 via CP030.


Overall I think this is a good step forward for Compound & Gauntlet. The protocol needs teams to analyze the protocol and provide input to improve the market further. Compound is a protocol of risk management. Without community members and users maintaining and improving the protocol, we won’t thrive.

At a high level, the proposal says Gauntlet will recommend collateral factors, borrow caps, reserve factors, and liquidation incentives based on their propriety simulations and a quarterly poll to determine the community’s risk tolerance.

Their key metric is Value at Risk or VaR.

The current Compound market is ~$8.5B borrowed, and there are 15 coins (excluding deprecated markets). That means we would pay Gauntlet 3,614 COMP or ~$1.67m for the quarter.

This isn’t a small amount of money to pay for an uncertain return. That being said, I think Gauntlet has demonstrated good intent. I will vote for this quarter, but I think the first quarter is a trial. I expect Gauntlet to deliver on all of the points below to secure a second quarter.

  • Build and maintain a dashboard that provides valuable insight and data.

  • Recommend collateral factors and provide sufficient evidence to back their recommendations.

  • Proactively support setting initial collateral for new assets.

  • Regularly interact with the community.


  1. What happens with new markets? Today we have 15, but if you get paid and then new markets get added, will you provide support for those?

  2. Do you have experience in recommending borrow caps & reserve factors? I didn’t see them mentioned in the latest Compound report.

  3. Does Compound have a point person at Gauntlet that community members can ask questions to?

  4. Do you have more info on how your Value at Risk statistic works? VaR isn’t mentioned in your most recent reports.

  5. What data are using as inputs for market liquidity and historical prices to calculate volatility?

Last point: The protocol can fund more than one team to work on this. If you read this and think your team could do risk analysis and parameter recommendation, please submit a proposal. There is enough work and money to go around!


New markets are immediately supported. Gauntlet’s pricing is only adjusted quarterly.

Gauntlet provides Acala (Karura) debt ceiling recommendations. The debt ceiling recommendation may not be directly applicable to borrow caps, but we will modify our methodology to account for the difference. We are also preparing to recommend borrow caps for Benqi and Aave once supported.

For reserve factors, we have formulated the optimization problem in our more recent Aave Market Risk Assessment (see Appendix C.6). The function design will hold well for Compound.

The community should consider me the default but can expect communications from many people on the Gauntlet team across Product and Data Science.

In our previous reports (both Aave and Compound), we focused on market risk and protocol resiliency. For example, in Aave, we used the key metrics of asset insolvency or safety module slashing to benchmark simulation runs. These numbers become difficult to compare to key metrics for yield and user borrow behavior metrics. In addition, the metrics can often be skewed for various market conditions. Our VaR metric will be a conflation of liquidation loss and insolvency metrics that can be quantified as some percentage of users’ capital that can make it easier to compare to upside (yield) metrics as well as scaled for tail market events.

Gauntlet ingests historical price and liquidity data for CEXes from Amberdata and for DEXes from onchain data.

1 Like

Is there a possibility of open-sourcing all of the risk monitoring tools, frameworks, and dashboards? So that in the next 1-2 years this can lead to a more decentralized risk team?

The goal of the Compound DAO should be to fund open-source teams and units that can benefit the protocol in the event any one person or company ceases to exist. I would even vote for the commission rate to be increased if the end goal is to develop open-source risk tools and frameworks that can work for any decentralized money market built on Ethereum.


Gauntlet will continue to formalize an Asset Onboarding Framework with the community and quantitatively clarify our methodology on request, as we have done previously.

To answer the question directly, Gauntlet currently has no plans to open-source our proprietary risk models and simulation SDK.

1 Like