[MIP 55/56] Risk Parameter Updates for 2023-06-11

Review of [MIP 55/56] Risk Parameter Updates for 2023-06-11

Summary

Gauntlet’s proposal aims to address three concerns:

  1. Reduce Multichain markets exposure on Apollo
  2. Optimize protocol revenues and lender returns for markets in which borrowing demand appears to be inelastic to rate on Artemis
  3. Allow more leverage for healthy markets on Artemis

We are in favor of the changes proposed to address all of the concerns, and have ensured that all parameters can be safely applied.

  1. By our assessment, the risk parameter updates for Apollo proposed by Gauntlet should be effective at reducing exposure to Multichain bridge assets. It is also our observation that liquidity levels for some markets (ETH.multi, USDC.multi, USDT.multi) do require some changes in borrow caps and/or collateral factor.
  2. We also confirm that borrow cap/collateral factor increases for xcUSDT and USDC.wh on Artemis can be safely applied.
  3. We are also in favor of applying the proposed interest rate model updates on Apollo and Artemis. In fact, a sizable portion of borrowing demand may be sufficiently inelastic to interest rate, as observed by Gauntlet. These changes will be an opportunity to measure real-world impact, and can be easily reverted in any case.

Outside of the proposed changes, we would also recommend to closely monitor WETH.wh and WBTC.wh liquidity levels over the next weeks and apply borrow caps / collateral factor updates if situation does not improve.

Key Observations

Liquidity levels for some markets that have not been addressed in this proposal have been dropping drastically in the last 30 days. Although some of these concerns have already been addressed in MIP-52, further updates may be required soon. This is especially true for WETH.wh and WBTC.wh.

Recommendations

  • Artemis
    • Closely monitor liquidity level for ETH.wh and WBTC.wh. Decrease collateral factors and borrow caps if drop persists.
  • Apollo
    • Increase the jump multiplier for FRAX from 2.5 to 3.175 to set it in line with other stable markets
    • Closely monitor liquidity level for ETH.multi, USDC.multi, USDT.multi and WBTC.multi. Decrease collateral factors and borrow caps if drop persists.

Robustness Analysis

Our analysis tests the robustness of all of the following changes proposed by Gauntlet:

Apollo
Symbol Parameter Current Recommended
ETH.multi Borrow Cap 700 110
ETH.multi Collateral Factor 64% 62%
USDC.multi Borrow Cap 9,727,000 2,200,000
USDT.multi Borrow Cap 600,000 250,000

USDC.multi IRM Current Recommended
Base 0.0 -
Kink 0.6 -
Multiplier 0.05 -
Jump Multiplier 2.5 3.175
Reserve Factor 0.15 -

USDT.multi IRM Current Recommended
Base 0.0 -
Kink 0.6 -
Multiplier 0.05 -
Jump Multiplier 2.5 3.175
Reserve Factor 0.15 -
Artemis
Symbol Parameter Current Recommended
xcUSDT Borrow Cap 1,200,000 1,300,000
xcUSDT Collateral Factor 53% 55%
USDC.wh Borrow Cap 2,600,000 3,000,000

WETH.wh IRM Param Current Recommended
Base 0.02 -
Kink 0.6 -
Multiplier 0.15 0.1875
Jump Multiplier 3 -
Reserve Factor 0.25 -

USDC.wh IRM Param Current Recommended
Base 0 -
Kink 0.8 -
Multiplier 0.05 0.0625
Jump Multiplier 2.5 -
Reserve Factor 0.15 -

xcUSDT IRM Param Current Recommended
Base 0 -
Kink 0.8 -
Multiplier 0.05 0.0625
Jump Multiplier 2.5 -
Reserve Factor 0.15 -

Apollo Governance Parameter Review

ETH.multi Borrow Cap Decrease

Symbol Parameter Current Recommended
ETH.multi Borrow cap 700 110

Test:
Protocol short exposure to ETH.multi is manageable

Result: :green_circle: Pass
Debt position of 20% of borrow cap could be theoretically liquidated

Details

Methodology:
All of the following trades can be executed with under 5% slippage

  • #1 Buy 20% of ETH.multi borrow cap from stables
  • #2 Liquidate the largest ETH.multi debt position

Results
Here’s the relevant market data concerning ETH.multi:

Test case #1 Slippage to purchase 20% of ETH.multi borrow cap:
20% of borrow cap = 22 ETH.multi ($38.2k)
:yellow_circle: Slippage to buy $38.2k ETH.multi = 19%
Note: Liquidation could be executed in chunks of $10k at <5% slippage

Tests case #2 Slippage to liquidate the largest ETH.multi debt position supposing stables collateral:
Largest debt position: $110k (0xe29a…7f6e)
:green_circle: Slippage to buy $110k ETH.multi = 0% (recursive strategy)
Note: Account is holding a recursive strategy which doesn’t pose significant liquidity risk as-is. However, if collateral was held in a different currency, the position would likely be very hard to liquidate.

ETH.multi Collateral Factor Decrease

By our assessment, the proposed collateral factor for ETH.multi should provide enough time for liquidators to clear currently existing risky loans given an extreme market downturn event.

Given a worst case scenario where top 5 non-recursive ETH.multi positions get liquidated, it may require more than 60 minutes for the liquidators to clear the risky loans given the current liquidation discount (10%). The proposed collateral factor does provide an extra buffer to compensate for the additional time that may be required for liquidators to profitably execute the liquidation.

Gauntlet’s recommendation:

Symbol Parameter Current Recommended
ETH.multi Collateral factor 64% 62%

Since the liquidation discount works together with the collateral ratio to provide a buffer for liquidations to be executed, we have first validated that the liquidation discount is sufficient to prevent bad debt from accumulating, then we’ve addressed the proposed collateral ratio change.


Test:
Liquidation discount test - Worst case historical liquidation scenario can be executed profitably in under 60 min.

Results: :yellow_circle: Partially pass
Passes given current market usage, but fails given worst case scenario

Details

Methodology: Validate that the worst time to liquidation is lower than 60 min by running a liquidation backtesting simulation.

Results
The largest amount of ETH.multi collateral that could be liquidated profitably within 60 minutes given the worst market conditions of the last 30 days is $11.5k. (see Liquidation backtesting results)

Considering the current distribution of liquidation prices of ETH.multi lenders and ETH volatility, this limit poses low risk for the protocol to accumulate bad debt. Most of the exposure comes from the $80.5k ETH.multi non-recursive collateral position held by 0xcf7b…c99d.

However, considering a worst case scenario where the top 5 non-recursive collateral positions (~$150k) get liquidated all at once, the probability of accumulating bad debt would be high. Liquidators would likely need to seize collateral and repay the debt in chunks of less than $10k, which would increase the total required time to clear all of the risky loans.

Top 5 non-recursive ETH.multi collateral positions


Test:
The proposed collateral factor gives the protocol sufficient room to wait for 60 min to execute a liquidations profitably without incurring bad debts even if the collateral assets decrease by the max drawdown

Results: :green_circle: Pass
Proposed collateral ratio provides 647 bps additional buffer over minimal requirement

Details

Methodology:
Validate that (1 - collateral ratio) covers the sum of following values at minimum:

  • Max drawdown for the duration of the worst case liquidation scenario (60min)
  • Liquidation incentive

Results
1h max drawdown (all time) = -21.53% (see ETH volatility)
Liquidation incentive = 10%
Total buffer requirement = 31.53%

Buffer provided by 62% collateral ratio = 38% (647 bps more than buffer requirement)


Test:
Parameter change doesn’t make any account liquidatable

Results: :green_circle: Pass
No liquidatable wallets

Details

Methodology:
Validate no account is liquidatable by running a collateral at risk simulation with following inputs:

  • Proposed collateral ratio for all markets.
  • Small asset prices decrease (-2%) across all collateral assets.

Run another simulation:

  • Proposed collateral ratio for all markets.
  • Small asset prices increase (+2%) across all debt assets.

Results
Simulation #1: 0 wallets at risk
Simulation #2: 0 wallets at risk


Test:
Collateral factor change do not increase market risk exposure beyond desired level

Results: :green_circle: Pass
Collateral at risk for all tested scenarios is not significantly impacted.

Details

Methodology:
Run Collateral at risk simulations with and without proposed parameter changes for the following scenarios:

  • #1: Collateral assets 5% historical VaR (excl. stables)
  • #2: Borrow assets 5% historical VaR (excl. stables)
  • #3: Stablecoin depeg

Make sure the risk exposure on the short and long side is within desired risk tolerance level if applicable.

Results:
#1 Collateral asset 5% historical VaR (excl stables)

  • Before: $27.9k Collateral at risk (74 wallets) (screenshot)
  • After: $27.9k Collateral at risk (74 wallets) (screenshot)

#2 Borrow assets 5% historical VaR (excl. stables)

  • Before: $3.7k Collateral at risk (60 wallets) (screenshot)
  • After: $3.7k Collateral at risk (60 wallets) (screenshot)

#3 Stablecoin depeg

  • Before: $3.3k Collateral at risk (43 wallets) (screenshot)
  • After: $3.3k Collateral at risk (48 wallets) (screenshot)

USDC.multi Borrow Cap Decrease

It is our assessment that the proposed borrow cap lowers the USDC.multi exposure for the protocol at a manageable level given recent outflows in liquidity.

In a worst case scenario where an account borrows 20% of the borrow cap in a highly leveraged strategy with other stables as collateral, the position would likely not accumulate bad debt given a downturn event.

Symbol Parameter Current Recommended
USDC.multi Borrow cap 9,727,000 2,200,000

Test:
Protocol short exposure to USDC.multi is manageable

Results: :green_circle: Pass
Debt position of 20% of borrow cap could be theoretically liquidated

Details

Methodology:
For every market, all of the following trades can be executed with under 5% slippage

  • #1 Buy 20% of USDC.multi borrow cap from stables
  • #2 Liquidate the largest USDC.multi debt position

Results

Here’s the relevant market data concerning USDC.multi:

Test case #1 Slippage to purchase 20% of USDC.multi borrow cap from stables:
20% of borrow cap = 440k USDC.multi ($440k)

  • :green_circle: Slippage to buy $440k USDC.multi = N/A (can’t be executed in a single swap)
  • Note: Swap could be executed in 2 chunks of $200k at <5% slippage

Tests case #2 Slippage to liquidate the largest USDC.multi debt position:

  • Largest debt position: $242k USDC.multi (0x6d98…e39e)
  • :green_circle: Slippage to buy $242k USDC.multi from stables = <5% (in 2 trades)
  • Note: Account is currently holding a recursive strategy which doesn’t pose significant liquidity risk as-is. However, if collateral was held in a different currency, the position may be very hard to liquidate.

USDT.multi Borrow Cap Decrease

It is our assessment that the proposed borrow cap lowers the USDT.multi exposure for the protocol at a manageable level given recent outflows in liquidity.

In a worst case scenario where an account borrows 20% of the borrow cap in a highly leveraged strategy with other stables as collateral, the position would likely not accumulate bad debt given a downturn event.

Symbol Parameter Current Recommended
USDT.multi Borrow cap 600,000 250,000

Test:
Protocol short exposure to USDT.multi is manageable

Results: :green_circle: Pass
Debt position of 20% of borrow cap could be theoretically liquidated

Details

Methodology:
For every market, all of the following trades can be executed with under 5% slippage

  • #1 Buy 20% of USDT.multi borrow cap from stables
  • #2 Liquidate the largest USDT.multi debt position

Results
Here’s the relevant market data concerning USDT.multi:

Test case #1 Slippage to purchase 20% of USDT.multi borrow cap:

  • 20% of borrow cap = 50k USDT.multi ($50k)
  • :green_circle: Slippage to buy $50k USDT.multi from collateral asset (MOVR) = 19%
  • Note: Could be executed in 5 swaps of ~$10k at <5% slippage.

Test case #2 Slippage to liquidate the largest USDT.multi debt position:

  • Largest debt position: $37k USDT.multi (0xcf7b…c99d)
  • :green_circle: Slippage to buy $37k USDT.multi from collateral asset (ETH.multi) = 31%
  • Note: Trade could be executed in chunks

USDC.multi and USDT.multi IRM updates

Symbol Parameter Current Recommended
USDC.multi / USDT.multi Jump Multiplier 2.5 3.175

We are in favor of a higher Jump Multiplier for USDC.multi and USDT.multi and support Gauntlet’s proposal. We would even advocate for the Jump Multiplier to be higher than 3.175 in order to make it more penalizing for borrowers when utilization is above the kink. Given the recent concerns regarding multichain assets and utilization momentarily being above the kink, we think this change is appropriate.

Moreover, we would also be in favor of increasing the Jump Multiplier for FRAX. We think such a change for FRAX would be prudent and would place all stablecoins on an equal footing.

Artemis Governance Parameter Review

xcUSDT Borrow Cap Increase

Increasing xcUSDT borrow cap is safe by our assessment.

Symbol Parameter Current Recommended
xcUSDT Borrow cap 1,200,000 1,300,000

Test:
Protocol short exposure to underlying assets is manageable

Results: :green_circle: Pass
Debt position of 20% of borrow cap could be theoretically liquidated

Details

Methodology:
For every market, all of the following trades can be executed with under 5% slippage:

  • #1 Buy 20% of xcUSDT borrow cap from stables
  • #2 Liquidate the largest xcUSDT debt position

Results
Here’s the relevant market data concerning xcUSDT:

Test case #1 Slippage to purchase 20% of xcUSDT borrow cap:

  • 20% of borrow cap = 260,000 xcUSDT ($260k)
  • :green_circle: Slippage to buy $260k xcUSDT from stables = 0.12%

Tests case #2 Slippage to liquidate the largest xcUSDT debt position:

  • Largest debt position: $128.8k (0xb554…1dab) \
  • :green_circle: Slippage to buy $128.8k xcUSDT from position collateral (xcUSDT) = 0%
  • Note: Account is holding a recursive strategy which doesn’t pose significant liquidity risk as-is. However, if collateral was held in a different currency, the position may be very hard to liquidate.

xcUSDT Collateral Factor Increase

Increasing xcUSDT collateral factor is safe by our assessment.

Symbol Parameter Current Recommended
xcUSDT Collateral Factor 53% 55%

Test:
Worst case historical liquidation scenario can be executed profitably in under 60 min.

Results: :green_circle: Pass
Could profitably liquidate $500’000 xcUSDT collateral in under 60min at any moment during the last 30d.

Details

Methodology:
Validate that top 5 xcUSDT collateral positions could have been liquidated together profitably in under 60min at all time during the last 30d using the liquidation backtesting tool:

  • Liquidation amount: Sum of top 5 xcUSDT collateral positions
  • Time period: last 30 days
  • Asset: xcUSDT
  • Conditional price change: 0
  • Liquidation discount: 7% (10% - 300bps reserve fee)

Results
Top 5 xcUSDT collateral positions = ~$500k:

Worst time to liquidation: 0 minutes (debt can be cleared instantly at any given moment)


Test:
The proposed collateral factor gives the protocol sufficient room to wait for 60 min to execute a liquidations profitably without incurring bad debt even if the collateral assets decrease by the max drawdown

Results: :green_circle: Pass
Proposed collateral ratio provides 3,128 bps additional buffer over minimal requirement

Details

Methodology:
Validate that (1 - collateral ratio) covers the sum of following values at minimum:

  • Max drawdown for the duration of the worst case liquidation scenario (60min)
  • Liquidation incentive

Results
1h max drawdown (all time) = -3.72% (see USDT volatility)
Liquidation incentive = 10%
Total buffer requirement = 13.72%

Buffer provided by 55% collateral ratio = 45% (647 bps more than buffer requirement)


Test:
Parameter change doesn’t make any account liquidatable

Results: :green_circle: N/A
Collateral factor increase can’t make any account insolvent

USDC.wh Borrow Cap Increase

Increasing USDC.wh borrow cap is safe by our assessment.

Symbol Parameter Current Recommended
USDC.wh Borrow cap 2,600,000 3,000,000

Test:
Protocol short exposure to underlying assets is manageable

Results: :green_circle: Pass
Debt position of 20% of borrow cap could be theoretically liquidated

Details

Methodology:
For every market, all of the following trades can be executed with under 5% slippage

  • #1 Buy 20% of USDC.wh borrow cap from stables
  • #2 Liquidate the largest USDC.wh debt position

Results
Here’s the relevant market data concerning USDC.wh:

Test case #1 Slippage to purchase 20% of USDC.wh borrow cap:

  • 20% of borrow cap = 600’000 xcUSDT ($600k)
  • :green_circle: Slippage to buy $600k USDC.wh from stables = 5%

Tests case #2 Slippage to liquidate the largest USDC.wh debt position:

  • Largest debt position: $600k (0xf743…b140, recursive position)
  • :green_circle: Slippage to buy $600k USDC.wh from stables = 5%

ETH.wh IRM update

Symbol Parameter Current Recommended
ETH.wh Multiplier 0.15 0.1875

Our analysis supports Gauntlet’s proposal to increase ETH.wh Multiplier from 0.15 to 0.1875 under the premise that ETH.wh borrow demand is sufficiently inelastic to increase the market’s borrow interest rate. This relatively small change has the potential to generate additional reserve fees over time.

USDC.wh IRM update

Symbol Parameter Current Recommended
USDC.wh Multiplier 0.05 0.0625

Given the current high utilization of USDC.wh we do think that an increase in the multiplier from 0.05 to 0.0625 is adequate and will likely generate overall higher reserve fees for the protocol. If utilization decreases substantially due to this change the multiplier could be lowered again if necessary.

xcUSDT IRM update

Symbol Parameter Current Recommended
xcUSDT Multiplier 0.05 0.0625

Since the xcUSDT utilization has been hovering close to the Kink for the past 30 days we do think that an increase in the multiplier from 0.05 to 0.0625 is adequate and will likely generate higher reserve fees for the protocol. If utilization decreases substantially as a consequence of this change the multiplier could be lowered again if necessary. This change is in line with our proposed recommendation in MIP-50.

2 Likes