[MIP 29/30] Risk Parameter Updates (2023-02-20)

Apollo Summary

Gauntlet recommends the following parameter changes:

  • Increase USDC.multi collateral factor from 66.0% to 68.0%.
  • Increase the ETH.multi collateral factor from 62.0% to 64.0%.
  • Increase the USDT.multi collateral factor from 44.0% to 46.0%.

The VaR is $0 and our recommendations will leave it unchanged. The LaR is $369k and our recommendations will leave it unchanged.

  • USDC.multi has a VaR of $0 and a LaR of $284k.
  • USDT.multi has a VaR of $0 and a LaR of $26.
  • WMOVR has a VaR of $0 and a LaR of $55k.
  • xcKSM has a VaR of $0 and a LaR of $29k.
  • ETH.multi and FRAX both have a VaR and LaR of $0.

ETH.multi, USDC.multi, and USDT.multi are relatively safe from a market risk perspective, so can have their collateral factors gradually increased to improve capital efficiency. FRAX, WMOVR, and xcKSM’s collateral factors are effectively balancing risk and capital efficiency.

Methodology

This set of parameter updates seeks to maintain the overall risk tolerance of the protocol while making risk trade-offs between specific assets.

Gauntlet’s parameter recommendations are driven by an optimization function that balances 3 core metrics: insolvencies, liquidations, and borrow usage. Parameter recommendations seek to optimize for this objective function. Our agent-based simulations use a wide array of varied input data that changes on a daily basis (including but not limited to asset volatility, asset correlation, asset collateral usage, DEX / CEX liquidity, trading volume, expected market impact of trades, and liquidator behavior). Gauntlet’s simulations tease out complex relationships between these inputs that cannot be simply expressed as heuristics. As such, the input metrics we show below can help understand why some of the param recs have been made but should not be taken as the only reason for recommendation. To learn more about our methodologies, please see the Helpful Links section at the bottom.

Supporting Data

The below figures show trends on key market statistics regarding borrows and utilization that we will continue to monitor:

Top 10 Borrowers’ Aggregate Positions & Borrow Usages

Top 10 Borrowers’ Entire Supply

Top 10 Borrowers’ Entire Borrows

Utilization Rate Over Time

Borrow Cap Utilization

Risk Dashboard

The community should use Gauntlet’s Apollo Risk Dashboard to better understand the updated parameter suggestions and general market risk in Apollo.

Value at Risk represents the 95th percentile insolvency value that occurs from simulations we run over a range of volatilities to approximate a tail event.

Liquidations at Risk represents the 95th percentile liquidation volume that occurs from simulations we run over a range of volatilities to approximate a tail event.

Artemis Summary

Gauntlet recommends the following parameter changes:

  • Increase USDC.wh collateral factor from 40.0% to 50.0%.
  • Increase WBTC.wh collateral factor from 25.0% to 30.0%.
  • Increase WETH.wh collateral factor from 35.0% to 45.0%.
  • Increase xcDOT collateral factor from 62.0% to 64.0%.
  • Increase WETH.wh borrow cap from 400 to 800
  • Increase WBTC.wh borrow cap from 30 to 60
  • Increase USDC.wh borrow cap from 516,000 to 1,000,000

The VaR is $358k and our recommendations will increase it by $2k to $360k. The LaR is $58k and our recommendations will increase it by $3k to $61k.

  • FRAX has a VaR of $360k and a LaR of $47k.
  • USDC.wh has a VaR of $169 and a LaR of $11.9k.
  • WETH.wh has a VaR of $326 and a LaR of $0.
  • xcDOT has a VaR of $3.76k and a LaR of $46k.

USDC.wh, WBTC.wh, WETH.wh, and xcDOT are relatively safe from a market risk perspective, so their collateral factors can be gradually increased to improve capital efficiency. BUSD.wh, FRAX, and WGLMR’s collateral factors are effectively balancing risk and capital efficiency.

Borrow Amounts for USDC.wh, WETH.wh, and WBTC.wh have approached near full utilization of their respective borrow caps. Gauntlet analyzed the assets’ liquidity on-chain and potential risk of short market manipulation attack to assess the risk of increasing borrow caps on the protocol. Gauntlet recommends increasing the borrow cap for these assets in order to optimize user experience and growth while mitigating tail risks to the protocol.

BUSD Increased Risk

Gauntlet is reaching out to the Artemis community to address the recent news related to the regulatory actions against BUSD and Paxos. As of recent news reports, Paxos plans to sunset BUSD but they plan to maintain it fully backed and redeemable until February 2024. Binance plans to support BUSD until they are able to find an alternative stablecoin.

However, Paxos has announced that they will cease minting the token as of February 21, 2023, which will lead to a decline in the circulating token supply of BUSD over time. Here are some of the following risk concerns we have for Artemis:

  • The liquidity of BUSD in Moonbeam has been gradually declining. On Stellaswap, the BUSD market’s liquidity has decreased by 30% from $1.2M to $853k (as of Feb 19, 2023) since the regulatory announcement against BUSD. Despite this, there is still a sufficient amount of liquidity in StellaSwap to support Artemis positions that are utilizing $158k BUSD as collateral and $276k BUSD in outstanding borrows. However, if the decline in BUSD liquidity on Stellaswap continues, it could result in higher slippage and increase the likelihood of insolvencies.

  • As it currently stands, 99.73% of the BUSD on Moonbeam is concentrated between StellaSwap and Artemis. However, if the liquidity of BUSD on StellaSwap continues to decrease, Artemis will surpass StellaSwap as the largest holder of BUSD on the network in relation to the circulating supply and could potentially lead to overweight exposure to the asset.

  • Furthermore, there have been multiple publications questioning whether there are sufficient reserves to support BUSD, this could potentially lead to market uncertainty about BUSD’s reserves and cause the asset to depeg from the US dollar across all markets.

BUSD Potential Next Steps

Given the increased risk, Gauntlet is recommending a deprecation plan should the communituy decide to deprecate the Artemis BUSD market. This plan entails reducing exposure to BUSD by using the listed risk parameters.

  • Freeze the BUSD market. Pause both borrow and supply.

    • This will prevent users from opening any further positions in the liquidity pool.
  • Adjust the IR curves to disincentivize further supply and borrow of BUSD

  • Increase reserve factor to 99% for BUSD (suppliers will stop receiving interest)

    • This will increase BUSD reserves from $6 to ~$37 per day and incentivize suppliers to remove their BUSD positions since they are not receiving any further Supply APY.
  • Incrementally reduce collateral factors every 2 weeks, shown in the following table:

Rec Cadence Initial 2 weeks 4 weeks 6 weeks
CF Reduction .2 → .15 .15→.1 .1→.05 .05→0
Liquidation ($) Impact 0 0 1 4
Liquidation Accounts Total Supply Impact 0 0 $37,630.95 $467,561.28
Estimated Active Collateral Usage ($) $118.5k $78.6k $39k $0k
Estimated Loss Annualized Reserve Impact -$612.64 -$618.84 -$614.19 -$605

Gauntlet will continue to monitor BUSD news and liquidity positions to ensure an efficient deprecation. As BUSD CF is reduced, we will update our deprecation plan as needed. Depending on developing BUSD market behavior and regulatory risk, certain adjustments may need to be expedited.

Methodology

This set of parameter updates seeks to maintain the overall risk tolerance of the protocol while making risk trade-offs between specific assets.

Gauntlet’s parameter recommendations are driven by an optimization function that balances 3 core metrics: insolvencies, liquidations, and borrow usage. Parameter recommendations seek to optimize for this objective function. Our agent-based simulations use a wide array of varied input data that changes on a daily basis (including but not limited to asset volatility, asset correlation, asset collateral usage, DEX / CEX liquidity, trading volume, expected market impact of trades, and liquidator behavior). Gauntlet’s simulations tease out complex relationships between these inputs that cannot be simply expressed as heuristics. As such, the input metrics we show below can help understand why some of the param recs have been made but should not be taken as the only reason for recommendation. To learn more about our methodologies, please see the Helpful Links section at the bottom.

Supporting Data

The below figures show trends on key market statistics regarding borrows and utilization that we will continue to monitor:

Top 10 Borrowers’ Aggregate Positions & Borrow Usages

This large BTC whale account 0xb4B7c91a46C7d4BFFe99D2ce4b056a5B280ed8e3 opened a borrowable position Feb 15th.

Top 10 Borrowers’ Entire Supply

Top 10 Borrowers’ Entire Borrows

Utilization Rate Over Time

Borrow Cap Utilization

Individual Account Analysis

Gauntlet wants to notify the community about account 0xb4B7c91a46C7d4BFFe99D2ce4b056a5B280ed8e3 opening an borrow position on Moonwell. This account has a large position representing 99% of WBTC.wh supply on Artemis. The supply balance for this account represents 92% of the circulating supply of WBTC.wh.

Position on Artemis

Borrow Usage

The position is taking a small recursive position in Artemis. Our models and date do not estimate any impactful risk of insolvency.

Risk Dashboard

The community should use Gauntlet’s Gauntlet’s Artemis Risk Dashboard to better understand the updated parameter suggestions and general market risk in Artemis.

Value at Risk represents the 95th percentile insolvency value that occurs from simulations we run over a range of volatilities to approximate a tail event.

Liquidations at Risk represents the 95th percentile liquidation volume that occurs from simulations we run over a range of volatilities to approximate a tail event.

Quick Links

Please click below to learn about our methodologies:

Gauntlet Parameter Recommendation Methodology
Gauntlet Model Methodology

1 Like

Hi guys, as pointed on in the Discord, the proposal for MIP 30 appeared to have the intention of raising the Collateral Factor of xcDOT from 62% to 64%. However, the proposed (and executed) contract call is actually setting the Reserve Factor to 64% , as indicated by:
*_setReserveFactor(640000000000000000)

Reserve Factors otherwise have been standardized to 15% for Stables and 25% for Volatile assets.

A higher reserve factor will send more of the borrow interest to the reserves instead of lenders. While higher reserves is not inherently bad thing, this essentially widens the lend/borrow spread on assets by lowering the Lending rate, and could have impacts to attracting sufficient liquidity.

This appears to simply just be an oversight in the proposal, but happy to discuss rationale if this was indeed the intention of the proposal.

1 Like

Hi MikeP,

You are correct this is an oversight. Gauntlet will be submitting a proposal to correct this change tomorrow. We will be improving our proposal process immediately to avoid these issues in the future. Our apologies MW community, this won’t happen again.

2 Likes

Risk Parameter MIP 30 Update:

Please see below for a summary of the post-mortem we completed for this error.

Timeline:

  • 2023-03-20: MIP 30 was proposed for the community to vote on recommended Risk Parameter Updates. This proposal contained an error in the contract calls. The intention was to " Increase xcDOT collateral factor from 62.0% to 64.0%.". Instead, the proposal increased the reserve factor of xcDOT to 64%.
  • 2023-02-24: MIP 30 passed and later executed on chain.
  • 2023-02-27: MIP 31 was proposed to fix this error.
  • 2023-03-02: MIP 31 passed and later executed on chain.

Cause of Error:

  • A segment of the proposal GTM process includes a human review of the contract calls in each proposal. During this segment, the error was not identified.

Progress on Process Improvements and Preventing this in the Future:

To address our process gaps and avoid any future errors, we are

  • Adding more testing that will compare parameters to previous values as well as known good ranges to help catch the types of errors in the future

    • Specific type checking
    • Revamped safety score framework to aggressively close gaps
  • Renovating our processes to remove manual steps that add risk by adding more automated safety testing, including:

    • Migrating all parameter specifications to their human interpretable form, the same as used in forum posts and community comms
    • Generating a summary of changed parameters from the JSON spec, so there is a single source of truth
  • Continuing to increase code review coverage and working closely with the community to ensure testing is enabled and complete, including: joint proposal review with the Moonwell community and collaboration on tooling.

    • While automated tooling and checks are being developed, additional layers of safety testing are in place to ensure gaps are closed immediately.
1 Like