[ARC] - Chaos Labs Risk Parameter Updates - Aave V3 Avalanche - 2023.02.07

[ChaosLabs]

Simple Summary

A proposal to adjust nine (9) total risk parameters, including Loan-to-Value, Liquidation Threshold and Liquidation Bonus, across four (4) Aave V3 Avalanche assets.

Motivation

Chaos Labs’ Parameter Recommendation Platform runs hundreds of thousands of agent-based off-chain and on-chain simulations to examine how different configurations of Aave V3 risk parameters would behave under adverse market conditions - and find the optimal values to maximize protocol borrow usage while minimizing losses from liquidations and bad debt.

Note: As a general guideline, we limit the search range to ±3% change for all parameters as a high/low bound for a given proposal. This ensures more controlled changes and allows us to analyze their effect on user behavior before recommending further amendments to the parameters in cases the optimal configuration is outside this range.

It is important to note that the parameters across the different assets were not jointly optimized, meaning that for each asset, the risk parameter optimization was done against the current state of the protocol. After optimizing the configuration for each asset individually, we verify that the combination of parameter updates leads to improving the risk-reward balance by simulating the recommended changes together.

Please find more information on the parameter recommendation methodology here.

You can also view the simulation results and breakdown for the different assets by clicking on them on this page.

Increase LTs + LTVs (WAVAX, BTC.b, and LINK.e)

The output of our simulations reveals an opportunity to increase LTVs and LTs for WAVAX, BTC.b, and LINK.e, resulting in improved capital efficiency of the system, with a minor effect on the projected VaR (95th percentile of the protocol losses that will be accrued due to bad debt from under-collateralized accounts over 24 hours) and EVaR (Extreme VaR, the 99th percentile of the protocol losses that will be accrued due to bad debt from under-collateralized accounts over 24 hours)

Given the changes, we project an increase of ~0.9% in borrow usage across the assets, while the increase in VaR is <10K$. The simulation shows Extreme VaR to remain within the same level.

WAVAX

BTC.b

LINK.e

Decrease LT + LTV (WETH.e)

The simulations show that the LT and LTV set for WETH.e are not optimal and should be reduced to improve the overall safety of the protocol. Given the change, the projected decrease in borrow usage for the asset is 1.37%, while the decrease in VaR and EVaR is ~$7K and ~$120K, respectively.

As of 02/06/2023, reducing the LT by 3% will lead to no new liquidated accounts (link).

Community Risk Appetite

When considering the reduction of LTs and LTVs of an asset, there are more considerations to take into account that are challenging to quantify and are not currently accounted for in our simulations. These include, for example, the impact of forced liquidations on the public trust of the protocol, borrower user experience, and capital efficiency vs. competing protocols.

Given the risk-reward tradeoff presented above and other subjective considerations, we will present two options in the Snapshot vote - YAE/NAE for the LT+LTV reduction of WETH.e.

We are eager to hear the community’s risk preferences and invite a discussion regarding key voting considerations.

Recommendations

Asset	Parameter	Current	Recommended	Change
WAVAX	Liquidation Threshold	70%	73%	+3%
WAVAX	Loan-to-Value	65%	68%	+3%
BTC.b	Liquidation Threshold	75%	78%	+3%
BTC.b	Loan-to-Value	70%	73%	+3%
BTC.b	Liquidation Penalty	6.5%	5.5%	-1%
LINK.e	Liquidation Threshold	65%	68%	+3%
LINK.e	Loan-to-Value	50%	53%	+3%

As mentioned above, Chaos Labs will present two options in the Snapshot vote - YAE/NAE for the parameter updates of WETH.e shown below:

Asset	Parameter	Current	Recommended	Change
WETH.e	Liquidation Threshold	82.5%	79.5%	-3%
WETH.e	Loan-to-Value	80%	77%	-3%

Next Steps

We invite a community discussion around the proposed parameter updates and will follow up with a Snapshot vote in 5 days, barring community objection.

[Pauljlei]

Thanks, @ChaosLabs. Our risk analysis did not yield a strong view on the proposed LINK or WAVAX changes. However, we did want to flag the changes for BTC.b and WETH.e. Gauntlet cautions against increasing LT/LTV for BTC.b and decreasing the LT/LTV for WETH.e.

Gauntlet Recommendations

Gauntlet’s analysis disagrees with increasing LT/LTV for BTC.b and decreasing the LT/LTV for WETH.e. Given that in this instance, the current supply for BTC.b (3700) has considerably exceeded both Gauntlet’s conservative (900) and aggressive caps (2800), increasing the LT for BTC.b and decreasing the LT for WETH.e adds risk to less liquid / whale-concentrated assets, decreases the borrowing power of safer and more liquid assets on Avalanche, and has negligible impact on increased borrows and protocol revenue.

1. In the event borrow usage increases disproportionately, the onchain liquidity for these collateral assets may not be able to support their liquidation potential.
2. These assets’ borrow usage suggests these LT changes will have a negligible impact on protocol revenue (~$925).

Analyzing onchain liquidity provides insight into the risks here. Suppose borrow usage with these assets as collateral actually increases disproportionately (a tail risk that we aim to protect against). Consider the event in which the largest suppliers for each asset decide to maximally borrow against their collateral (which is a possible, permissible action in the protocol). Essentially, we aim to evaluate the ‘potential harm’ this user may cause. Consider the slippage that results from a liquidation sized at this max borrow, adjusted for the initial close factor of 50%.

• BTC.b, largest supplier 0xd275e5cb559d6dc236a5f8002a5f0b4c8e610701
• $28.6mm → max borrow, adjusted for close factor is ~$10mm
• (Also note that 60% of BTC.b supply is on Aave)

• WETH, largest supplier $9mm → max borrow, adjusted for close factor is ~$3.3mm

As shown above, the slippage for BTC.b is meaningfully higher than WETH.e for the respective largest suppliers.

Even though the marginal slippage corresponding to respective LT increases of 3% of the current supply may be acceptably small, BTC.b has the latent potential to become a high VAR asset due to the strong whale presence. Moreover, 60% of the BTC.b circulating supply on Avalanche is supplied in Aave already.

The largest historical liquidations on Aave v3 Avalanche have been < $200k in size. Gauntlet is researching the distribution of liquidations that use on-chain liquidity vs. liquidations that use global liquidity. Until we find sufficient evidence to the contrary, we should not assume a larger base of liquidity (global liquidity) for potential large liquidations that have never historically happened before and should rely solely on on-chain liquidity.

Next, let’s analyze the borrow usage (taken from the Chaos dashboard, as of ~2/12/2023)-

• WETH.e has borrow usage of 46%, BTC.b has 37.8%, LINK.e has 26.5%, WAVAX has 13.8%.

After adjusting for asset CF -

• WETH.e has borrow usage of 47%, BTC.b has 45%, LINK.e has 34%, WAVAX has 15%.

Some of the supplied collateral is unproductive, as in changes to the LT should not rationally impact borrower behavior -

• For instance, roughly 30% of BTC.b borrow is recursive - any additional BTC.b borrow with BTC.b supply is not profitable relative to nonrecursive borrowing. So rationally, this BTC.b supply would not have new associated borrow and is thus non productive.

Assuming users will maintain their (debt / max debt) ratio, where max debt is the maximum debt they can take on with their current supply, we find that the BTC.B LT change could increase borrows by roughly $900k, while the WETH.e LT change decreases borrows by $525k. This is the breakdown by debt (borrow) asset when aggregated together -

which is roughly $370k increase in borrows. At current variable borrow rates of roughly 2.5%, with a 10% RF, this amounts to ~$925 in added protocol revenue per year.

TLDR;

Ultimately, increasing LT of BTC.b while decreasing the LT of WETH.e enables excess risk to less liquid / whale-concentrated assets and only has a small potential benefit on protocol revenue.

[ChaosLabs]

Thanks, @Pauljlei, for sharing Gauntlet’s detailed analysis.

This analysis highlights an important topic: assumptions around future user behavior and risk amongst markets that are “whale-concentrated.”

We believe that “whale-centric” liquidity distribution across different networks is common enough to warrant a discussion regarding the community’s appetite for risk.

We are proponents of selecting the risk-off approach in any dilemma while driving discussions that solicit community feedback. Therefore, as the proposed changes for BTC.b and WETH.e in this batch of recommendations have a small impact on protocol revenue, we will omit these assets and revisit them after gauging the community’s risk appetite and assumption preferences.

We will move forward with a Snapshot vote to update the risk parameters for LINK.e and WAVAX, as presented in the table in the original post.

[ChaosLabs]

We have published a Snapshot for the community to vote on, starting in 24h.

Thank you in advance for your participation in the vote.

[ChaosLabs]

The Snapshot vote for this proposal has passed with 561K YAE votes.
As there is no on-chain vote for V3 Avalanche, this proposal will be executed via the Aave Guardians.