ARC: Q2 Dynamic Risk Parameters

[inkyamze]

Summary

A proposal to renew Gauntlet’s Dynamic Risk Parameters engagement for continuous market risk management to optimize yield, capital efficiency, and mitigate depositor losses.

Background

Since mid-August, Gauntlet has provided Dynamic Risk Parameters and delivered on all areas outlined in the original proposal. Gauntlet has published 7 ARCs, Snapshots, and AIPs on parameter recommendations (102 parameter updates across 21 assets), 2 Market Risk Monthly reviews, new weekly risk updates in Aave News, and launched the Aave Risk Dashboard. Gauntlet has invested heavily in the proper infrastructure including ETL data pipelines, agent-based simulation infrastructure, and financial models that make possible the parameter recommendations and front-end Risk Dashboard.

The below will outline how Gauntlet has been able to create value for the Aave ecosystem. We also describe specific ways the Gauntlet Platform will make an even more significant impact moving forward.

Impact of the Gauntlet Platform

How the Gauntlet Platform drives value to Lenders, Borrowers, and Aave stakeholders:

• Mitigating losses from liquidations and insolvencies - $6.7M of annual value
• Increase capital efficiency and yield for users - $16.4M of annual value
• Safely position Aave for sustainable growth

Mitigating losses from liquidations and insolvencies - $6.7M of annual value

Gauntlet’s parameter recommendations save Borrowers’ collateral lost to liquidators by safely lowering liquidation bonuses. When Borrowers’ collateral are liquidated, they lose an extra portion of their collateral to liquidators. This bonus incentivizes liquidators but often can be reduced without sacrificing protocol risk.

From our first liquidation bonus updates (2021-08-27) to present Gauntlet has saved Aave V2 protocol users $218,141 in liquidations due to lower liquidation bonuses, which equates to $1,447,663 on an annualized basis. It is important to note that this time frame experienced relatively low levels of asset volatility. Higher levels of asset volatility would lead to more liquidations, and more savings realized.

Additionally, Gauntlet has reduced the total capital at risk due to insolvencies and liquidations when markets are under duress (Value at Risk). Gauntlet’s parameter recommendations reduce VaR, following the community’s risk tolerance, which protects Borrowers, as less capital is expected to be lost during market shocks. In addition, lower VaR accrues value to Aave token holders, as insolvencies lead to payouts from Aave’s Safety Module.

On average, Gauntlet’s recommendations have shown a 14% reduction in VaR. Our simulations observe an average of 0.48% of Aave V2’s TVL being at risk on any given day. In the scenario where volatility reaches similar levels as Black Thursday, we expect to save Aave users an average of $5,258,511 in liquidations and insolvencies.

In the future, should the community assign a riskAdmin role, Gauntlet can achieve more precise adjustments and streamlined risk management. Compared to the time-intensive ARC, Snapshot, and AIP processes, our response time during catastrophic market shocks would return tangible benefits. One example of this is adjusting liquidation bonuses swiftly to prevent liquidation cascades.

Increasing capital efficiency and yield for users - $16.4M of annual value

One of the core functions of the Gauntlet Platform is increasing capital efficiency for borrowers while controlling insolvency risk. Gauntlet’s recommendations have led to higher borrow demand following lower collateralization ratios. On Sept. 4th, 6.1% of users had collateralization ratios below 2.00. As of Oct. 21st, the number of accounts has grown by roughly 14%, and 8.9% of new accounts have collateralization ratios below 2.00. New positions opened after Gauntlet’s parameter recommendations have higher borrow demand, as seen in the histogram below.

Histogram of new accounts since 09/04 by Collateralization Ratio and Borrow Value

From another angle, Borrow Usage is an additional measure of capital efficiency and signifies how aggressively depositors of collateral borrow against their supply. Since Gauntlet’s parameter recommendations, we have seen a 1.1 percentage points increase in borrow usage.

Also, corresponding to these updates, an increase in yield for depositors has been realized. Depositor income the month before our parameter updates was $24.7M. The average monthly yield after Gauntlet’s parameter updates was $37.7M. Increased capital efficiency and the resulting benefit to depositors should, at least in part, be attributed to risk parameter optimizations. Given the increase in borrow usage for the same amount of assets lent, Depositors will earn an estimated $5.6M additional yield.

Let’s look at an illustrative example of how a fast path, via riskAdmin, to parameter changes can facilitate opportunistically increasing LTVs and magnify impact. Say borrowers on AAVE have at least a 2% return on their borrowed capital above borrow cost. If we assume 50% of the year is low volatility, Gauntlet raises LTVs by an average of 5% during low volatility periods, and ~$20B of collateral on Aave, then we can conservatively estimate $10M of value to Borrowers. Those new borrows would also generate over $800K worth of additional interest income for depositors.

Safely position Aave for sustainable growth

Gauntlet’s risk management empowers Aave not only to optimize existing assets but immediately support new assets. We have already made parameter updates for the newly added DPI in AIP-43.

Gauntlet is also closely monitoring the growth and stability of new assets FEI and FRAX. Should market conditions allow for safely turning on collateral, which initial simulations suggest can be done, we will seek community consensus to do so. Doing so could unlock $80M in collateral today and scaling proportionally, using the DAI borrowed on Aave to that in circulation, tens of millions more.

Safely supporting new collateral helps make the DAO sustainable beyond the Ecosystem Reserve.

Risk Dashboard

Gauntlet has launched a Risk Dashboard for the community to provide key insights into risk and capital efficiency. The dashboard focuses on both the system-level risk in Aave V2 and the market risk on an individual collateral level. Our goal is to help convey our methodology to the community and provide visibility into specific parameter recommendations. The dashboard will continue to be updated daily.

Expectations

Gauntlet will continue to deliver on the same risk parameters and communications plan outlined in the original Proposal while expanding the scope to include the two changes below:

• Configuring Reserve Factors
• Supporting Borrow Caps when the protocol is upgraded

Updated Cost

Gauntlet’s standard liquidity protocol fee model has been updated universally since the original proposal. The updated formula to calculate the service fee still has four components:

1. An asset multiplier to track risk management complexity
2. [Updated] A proxy for capital efficiency
3. [Updated] A marginal base fee
4. [Updated] 30-Day VWAP (Volume Weighted Average Price) of AAVE

Component #1 remains the same. The asset multiplier calculation is log(Number of Assets, 10)*. New assets on the protocol add complexity to risk management.

Components #2, #3, and #4 have been updated. Total Borrow remains the best proxy for capital efficiency and is now calculated as the 30-day average and rounded down to the nearest $1B. The base fee now takes marginal reductions to be more scalable for DAOs. Lastly, the requested AAVE denomination has been updated to protect the DAO and Gauntlet from short-term price fluctuations.

Gauntlet’s method for token transfer has also been standardized and improved. No longer will any portion of the requested payment, stkAAVE, not be subject to vesting. A Sablier stream will be requested for the full payment vesting over one calendar quarter.

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

^{*Gauntlet quarterly service fee denominated in AAVE (table above calculated at $312)}

---

^{*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 < $2b, there is no basis point fee. The formula is log(Assets,10) * $1,200,000 / 4 )}

Conclusion

In just a few months, the Gauntlet Platform delivered significant value to the Aave ecosystem, and importantly, the future impact will be even greater. Through rigorous quantitative analysis and agent-based simulation frameworks derived from industries including algorithmic trading and autonomous vehicles, the Gauntlet Platform empowers Aave to optimize yield, capital efficiency, and mitigate depositor losses. Gauntlet currently provides parameter recommendations covering over $49B of assets. Managing risk for numerous protocols allows our data scientists and engineers to improve our models ahead of realization events occurring on Aave. As Aave continues to scale, it is crucial to adapt and translate complexity into trade-offs to support community decisions.

About Gauntlet

Gauntlet is a simulation platform for market risk management and protocol optimization. Gauntlet’s continuous risk management work includes Compound, Acala, Benqi, and of course Aave. Gauntlet’s continuous incentive optimization work includes Balancer and SushiSwap.

Thanks to @tarun, @jmo, @pauljlei, @wfu, @henryprior, @htkao, @shaan, and many others for assistance on this ARC.

[inkyamze]

Next Steps

• Consensus check via Snapshot on 2021-11-02
• AIP following a successful Snapshot on 2021-11-09

[JeremyKeating]

It will be a yes vote from me. While the Gauntlet fee is a lot of money they are producing good results and creating value above their fee. My hope is that this hasn’t been initial easy wins but instead the kind of value that will be delivered from the relationship long-term. Hopefully we will develop tools to help in independently validating these benefits.
@inkyamze I’ve got just one suggestions for the proposal. Could you make clear roughly what you expect the fee to be. It is a little tricky to work out from the table. I think I know but I am not sure. Sorry if I have missed it somewhere. Saying something like x million USD for y months would be really helpful. I know it would be subject to caveats due to volatility and expansions etc but a rough estimate would be really helpful

[inkyamze]

Thanks for the support Jeremy!

We will recalculate the cost the day prior to any AIP but right now the metrics for Aave V2 are:

• 30-day average Total Borrow: $7.28B
• Assets: 31

Using the table above (AAVE price $312) the fee for this quarter would be 7170 AAVE ($2.237M). We just reran the 30-Day AAVE VWAP price and it is currently $349.55 which would be 6,400 AAVE (still $2.237M).

[inkyamze]

The consensus check has been created > Snapshot

[MarcZeller]

Hello,

Gauntlet is doing an awesome job.

Do this job has value? Yes without any doubt.

Do this job is worth 2M+$? I’ll argue that no and I voted no on the snapshot proposal.

The ecosystem reserve is not infinite and should be spent carefully.

Happy to vote yes on a more reasonable amount.

One pathway for more sustainable pricing is maybe to consider pricing in relation to what the fee collector contract actually collects.

[inkyamze]

The updated formula inputs will be used for the forthcoming AIP:

• 30-day average Total Borrow: $6.743443866457782E9 billion
• 30-Day AAVE VWAP: $ 345.231184773

[MarcZeller]

After AIP-46 rejection, it seems clear that this proposal does not reach a wider approval.

I think while none voiced concerns about the quality of Gauntlet’s work, concerns about the pricing model have been expressed during this vote, and discussion.

Work on an updated proposal that the community might favor enough in votes.

[jmo]

Taking over here for Gauntlet as Nick is out of office for the next few months.

Marc - there were only 5 no votes, which did not comprise a majority by weight or by count:

If this is a clear rejection, I am worried about what constitutes a clear endorsement . As we mentioned on Twitter, we know that some people had issues voting with this proposal. It sounds like you are interested in a new proposal - so are we. We use standard pricing across all of our clients, but your feedback on price is duly noted. We’re looking forward to providing more info early next week on how we can proceed.

[eboado]

I voted NO for the proposal. My rationale goes in line with @MarcZeller 's presented arguments:

• I think Gauntlet’s work, both the one done on the first quarter of collaboration with the DAO and previous collaborations with the Aave ecosystem is high-quality, and I am in favour of continuing the collaboration.
• On the other hand, I am a firm believer on self-sustainability of the ecosystem. Taking into account the cost of first quarter and the one presented in this proposal, we are talking about an approximated yearly cost an amount realistically between $6-10m. The Aave Treasury for the V2 market (which in my opinion should be considered as the main source of “operative” costs of the market), has at the moment ~$22m collected. Considering the current growth, this amount could be around $25m at the end of the year. We are talking about 30% of the total funds collected since inception to the treasury going exclusively on this specific collaboration with Gauntlet.
• More plain, I think ~$1.8m for a quarter of collaboration on this scope is just too high.

I’m more than open to change my vote to YES, but for that the terms should be changed via previous discussion with the broad community.

[jmo]

Achieving clarity on Risk Management for AAVE

We are going to move forward with a second vote that will allow for better participation and clearer next steps for Gauntlet, AAVE, and its users. We are targeting 11am PT, Tuesday 11/30 to begin the vote.

While the vast majority of users supported the proposal, we wanted to see if we could modify our deal structure to address the feedback we got from @eboado and @marczeller. As none of this feedback was negative regarding the quality of our work, we are proposing a discounted longer contract that will allow us to focus on adding value to AAVE.

New Pricing

We’ll use a similar but simplified pricing model that extends the contract renewal period to 1y, in turn for a 25% discount on our standard pricing. This results in a price of 7.5bps of total borrow per year. With a 30-day average Total Borrow of $6.74 billion and a 30-Day AAVE TWAP of $308.23, this sizes the AAVE grant at 16,412 AAVE. This provides AAVE with dynamic risk parameterization and a clear risk manager for customers through December 3rd, 2022.

One thing we’d like to reiterate is that Gauntlet’s fee is paid in the AAVE token, with vesting via Sablier. This does not impact the AAVE reserves and for comparison, is less than 2% of the yearly AAVE distribution for liquidity mining. Gauntlet has never sold any AAVE tokens.

• Like liquidity mining, Gauntlet’s work to manage LTVs and capital efficiency in the protocol can drive growth. Our token grant is tied to the amount of assets in the protocol for this very reason - we want to ensure that the fee we charge is tied to the value we deliver. As @inkymaze mentions above, starting with this renewal we are also supporting fee optimization in each market to ensure we drive as much value to AAVE token holders as possible. Doing some rough math here, a 10% increase in LTVs would result in close to $1B in excess borrow, $40mm of interest payments to AAVE users and $8mm in fees collected by token holders.

The Value We Provide

Fees collected by token holders are not the only way we can help AAVE and its users:

• We’ve gone above and beyond at every opportunity, including working with the response team for the recent issues with certain assets , pushing the proposal to disable assets that posed a threat to AAVE.
• As AAVE moves to V3, we’re here to help and work with the team to set reasonable values for parameters like Borrow caps that can help keep the protocol safe and enable the next era of growth
• In the end, working with Gauntlet is a promise to AAVE’s users that AAVE cares about risk and is working to provide the best capital efficiency that market conditions allow. This is just as important for retaining current users as attracting new ones; even more so for the latter as institutional capital continues to enter DeFi.

We have spent almost three years building our platform to measure and manage risk in DeFi and it is the cumulation of that multiyear effort that has resulted in the platform that AAVE depends on today. This work includes, but is by no means limited to:

• Heavily modifying geth so that we could run thousands of simulations a day with AAVE’s production smart contracts. This ensures key contributions to risk like liquidation spirals are modeled correctly.
• Incorporating seven different data providers into our platform to enable that we have accurate market information in our models for all collateral assets, even those with liquidity fragmented across multiple DEXs and CEXs.

---

We are looking forward to this second vote so we get a clearer signal from the community on how we should proceed. We’re happy to answer any questions people have here and hope that the changes we have made to the original proposal help address some of the concerns raised.

[Franklin]

On behalf of Pantera Capital, I’m strongly in favor of passing this updated proposal (and voted for the previous terms as well).

Disclosure: We do not have any financial interest in Gauntlet.

1) Risk management is a core need for Aave.

I’d argue that risk management is THE job for the Aave community. The “job to be done” in this particular case includes: quantifying risk, optimizing collateral parameters, running stress tests and raising the alarm when needed. These are the decisions that Wall Street failed to make in the 2008 financial crisis.

Aave’s top priority should be attracting the talent and skills needed to form a robust risk architecture for the long-term. That may consist of community initiatives like RiskDAO (still immature), specialized providers like Gauntlet and a collection of on-chain mechanisms still in early development / adoption (e.g., reactive governance, circuit breakers, auto-liquidation tools).

The service that Gauntlet provides today is one-of-a-kind. For Aave, this is a bug, not a feature.

Even if there’s a risk of overpaying, does that outweigh the risk of running a $10B+ lending protocol without a partner like Gauntlet? Does it even outweigh the benefits? Particularly the benefit of incentivizing new talent to either join Gauntlet or compete against them - either is great for the Aave community long-term.

2) It’s impossible to know whether you over-spent on risk management - only whether you under-spent.

Investing in risk management is similar to buying insurance or cybersecurity software. If nothing goes wrong, you’ll always feel like you over-spent. If something goes wrong, you’ll regret not spending more.

It’s difficult to gauge how much Aave should spend on risk management. As folks say, this is a new paradigm. Trying to copy and paste the playbook from traditional finance doesn’t make sense in a world run by smart contracts. It’s possible that the total $ spend on risk management in DeFi may be lower than TradFi or at least greatly diffused, but also that the % of overall protocol spend on risk management may be higher in DeFi than in TradFi.

50% seems high. 5% seems low. The updated proposal will cost 16k AAVE or ~$4-6m per year. As @jmo stated, that’s less than 2% of what Aave spends on liquidity mining (i.e., growth marketing). That seems really reasonable on a relative basis - even if it’s too early to judge on an objective basis.

I appreciate the spirit of financial stewardship in the Aave community, from @eboado and @MarcZeller. At the same time, financial stewardship isn’t only about managing costs down - it can also be about spending smartly by leaning forward in an area that seems crucially under-invested for Aave (and DeFi generally) and will pay dividends over the long-term.

Franklin @ Pantera Capital

[sakulstra]

Is any of the things(tooling not results) gauntlet provides open-source and can be continued / taken over by the community / risk dao at some point?

If the answer to this is no (i think it is?), I oppose this proposal as I fear it will lead into a sunken cost fallacy. 6M for “providing a service”, not “developing a service” is just a lot when once this engagement is not continued the community has to start at 0.

[Franklin]

I disagree with this viewpoint. There’s no particular reason why a service like this needs to be open-source or decentralized - as long as there’s open competition to be a selected provider. There should be multiple providers, in addition to RiskDAO’s output, and eventually a robust RFP process. Open competition is more important than open-source development in this case.

[MarcZeller]

This will be my last answer on the topic before going back to neutrality.

I’m a simple man using simple words and my feeling is that we’re trying to check the temperature with a pressure gauge while evaluating the budget dedicated to this.

The ecosystem reserve is a community-own resource that will run out eventually, it’s up to everyone’s responsibility in the Aave community to make sure this reserve is used at its best while it last. It’s a stock of money. There’s simply no sustainable way to make a budget of recurring expanses on a stock.

The protocol fees collected in the Aave collector contracts is the protocol revenue, it can go up and down with Aave protocol usage, but it’s a flow. Using this flow metric as the reference of the potential budget is the only way to build a sustainable relationship that can last for years and grow equally with the success of the protocol.

According to the Aave Weekly the protocol revenue for last week was 851k$ (44.3M$ annualized)

The current discounted proposal represents around 13% of ALL Aave protocol revenue.

so here are my last points on the topic:

1. let’s all agree to stop “total Borrow” bps metrics, the only sustainable way to establish a sustainable budget is to check the actual protocol revenue, and discuss based on this.
2. on this new sustainable metric, do we want as a community and as a protocol, each quarter, work two weeks for gauntlet and the rest for other stuff? do we want more? less? the clear consensus on discussions so far is that Gauntlet work is valuable and add value to the Aave protocol.
   but it’s up to the aave community to decide on this and I’ll go back to neutrality.
3. in reaction to @sakulstra, IMHO it’s all fair and square that Gauntlet doesn’t share tools that are highly valuable, anyone can build their own tools and “compete”.
4. it’s out of scope for this quarterly budget, but I’ll like to introduce the idea to gradually cover this budget less with AAVE assets from the ecosystem reserve et increasingly with stablecoins such as USDC from the aave collector. One day, the Ecosystem reserve will run out of funds, and we need to think ahead and gradually prepare for the transition.

[jmo]

Marc - we appreciate your thoughts here as always. Happy to let you have the last word here - we’ve already covered why we think the proposal is clearly ROI positive for AAVE and its community. Looking forward to seeing how other members of the community feel with this vote, which is now live:

Aave - Open Source Liquidity Protocol