Gauntlet Interest Rate Recommendations for WETH and WMATIC on v2 and v3

[Gauntlet]

Summary

Gauntlet’s analysis has identified opportunities to adjust interest rate parameters for WETH on Ethereum v3 and v2, Optimism v3, Arbitrum v3, Polygon v3, as well as WMATIC on Polygon v3 and v2. This will help to facilitate more Emode borrowing against LST collateral, which may

• drive additional revenue to Aave (estimated ~$1m, based on assumptions in the below Impact section)
• adjust current risk profile for LST to depend less on onchain liquidity, which has been steadily declining over the past few months. This may be adding risk to the current borrow composition against LST collateral.

Recommendations

Asset	Chain	Parameter	Current	Recommendation
WMATIC	Polygon v3	Variable Slope 1	0.061	0.043
WMATIC	Polygon v2	Variable Slope 1	0.06	0.043
WETH (*)	All v3	Variable Slope 1	0.038	0.033
WETH	All v3	Variable Base	0.01	0
WETH	Ethereum v2	Variable Slope 1	0.038	0.033
WETH	Ethereum v2	Variable BASE	0.01	0

(*) @MarcZeller made a recommendation to adjust WETH IR for v3 deployments here. Current parameters indicate that Variable Slope 1 is already 0.038, but the Variable Base is 0.01. We recommend lowering slope 1 even further, but not move Uopt.

Impact

Protocol	Symbol	Reserve Factor	Current Util	Current Borrow Rate	Current Supply Rate	New Borrow Rate	New Supply Rate	Current Borrow	New Borrow	Delta Borrow	Revenue Increase
Ethereum v2	WETH	0.2	0.530	0.035	0.015	0.023	0.010	3.88E+08	0	0	0
Arbitrum v3	WETH	0.15	0.517	0.035	0.015	0.023	0.010	3.56E+07	5.38E+07	1.82E+07	9.53E+04
Ethereum v3	WETH	0.15	0.492	0.033	0.014	0.022	0.009	2.84E+08	4.49E+08	1.66E+08	8.70E+05
Optimism v3	WETH	0.15	0.519	0.035	0.015	0.023	0.010	1.65E+07	2.48E+07	8.30E+06	4.36E+04
Polygon v3	WETH	0.15	0.250	0.022	0.005	0.011	0.002	1.08E+07	3.37E+07	2.29E+07	1.20E+05
Polygon v2	WMATIC	0.46	0.179	0.022	0.002	0.017	0.002	1.40E+06	0	0	0
Polygon v3	WMATIC	0.2	0.513	0.042	0.017	0.031	0.013	2.09E+07	2.97E+07	8.84E+06	7.96E+04

The above IR changes could drive new borrow rates at utilization levels just underneath the kink (rather than 20-30% away from the kink) to match current borrow rates, allowing for marginal increases in recursive WETH/WMATIC borrowing to remain profitable. Should that happen, the IR changes could generate ~$1m in additional annual revenue. We assume 0 revenue impact to v2 markets due to lack of Emode.

Analysis

LST collateralized BASE borrowing in Emode has been proportionally decreasing over the past few months, which could be lowering overall WETH and WMATIC borrows across Aave. This indicates that there is an opportunity to both increase revenue for Aave via encouraging more Emode borrows and also, from a risk perspective, to reduce current dependency on LST liquidity.

We welcome the community to follow our analysis and refer to our dashboard and appendix for more liquidity and interest rate visualizations.

• WETH and WMATIC have consistently had utilization below the kink.
  • Both assets have utilizations that live consistently around 50-60%, which is below the kink.
• Most of the LST collateral on these chains are supporting borrowing outside of Emode, a trend that has been increasing, as we detail in our appendix. This may be due to
  • Lower staking yields for LSTs. Current APRs for staking ETH on Lido (3.8%), RocketPool (3.27%), and Coinbase (3.27%) are below the current WETH borrow rate across v3.
  • The historical APR for staking ETH on Lido has been decreasing since May. Likewise, APY for staking ETH on Coinbase has been on the decline since July.
  • Likewise, the current APR for staking stMATIC is 4.3% on Lido and 4.82% for MaticX on Stader, compared ot the 4.2% borrow rate for WMATIC.
• LST collateral is supporting proportionally more borrow outside of Emode, and those borrows are primarily stablecoin based. As a result, recent improvements such as CSPA do not affect liquidation probability, which implies that LST onchain liquidity remains critical to ensuring healthy liquidations.
  • 
• LST liquidity has been declining over the past few months, which may be driven by declining LDO incentives. Lido used to incentivize LPing in Curve by distributing LDO as rewards, but this is no longer the case as of June.
  • Since the start of June, it seems like no rewards are being provided for LPs on any DEX, including Balancer. In fact, Dune shows that Balancer incentives ended July 2022.
  • The majority of the reduction in wstETH liquidity is due to decreases on Curve and Balancer, which seem to coincide with the removal of LDO incentives.

Next steps

Welcome community feedback and aim to put up snapshot on 2023-09-05.

Appendix

LST liquidity over time

wstETH on Arbitrum

LST APR Timeseries

Lido

Coinbase

WETH, WMATIC utilization rates

WMATIC historical utilization on Polygon v3

WETH historical utilization on Ethereum v3

Timeseries of Emode vs non-emode borrows against LST collateral

Decreasing LST yields is coupled with decreasing emode borrows against those LSTs.

Ethereum

Emode 1 is ETH-based emode.

cbETH

rETH

wstETH

Optimism

Emode 2 is ETH-based emode.

rETH

wstETH

Polygon

Emode 2 is MATIC-based emode.

stMATIC

MaticX

User behavior changes

Borrower behavior change after Slope 1 was reduced, Uopt was shifted on 2023.03.31, for Arbitrum v3 WETH.

[MarcZeller]

Hello, we disagree with these recommendations placing slope 1 much lower than staking APY yield (currently at 4.1% on stETH) as we believe it will lead to equilibrium above optimalRatio, decreasing overall protocol revenue and available liquidity.

In general With the ACI, we think there’s a real potential added value to track reference yield for LSTs (stETH lido yield) & (MaticX yield) and adjust the slope1 parameters algorithmically to a range targeting LST yield. There’s little governance added value to set these parameters and by definition, the DAO will always run behind the market.

It’s time to automate & optimize this.

We might propose a proposal implementing this in the coming weeks.

Meanwhile, we initially stand against this proposal. But we’re waiting for community & second risk team feedback to form a definitive opinion.

[TokenLogic]

Hi @Gauntlet,

Thank you for this proposal. As many know we work very closely with several teams building yield maximising strategies on Aave v3. We have even modelled some strategies to understand how they affect Aave Protocol in helping shape the strategies design.

We have a few concerns with the proposal in its current form.

• Polygon v2 is being deprecated and therefore the borrow rates should be increased relative to v3 over time to encourage migration.
• Polygon v3 whilst users who borrow wMATIC receive rewards, the borrow rates should not be reduced until after the rewards stop being distributed.
• Polygon LSTs yield maximising strategies receive incentives at the strategy level and with several integrations in the pipeline, we suggest not changing the rates until after we see the impact of this additional demand from new distribution channels.
• Setting Slope1 at less than the LST yield is okay in our opinion. However, it needs to take into consideration any incentives that affect the net borrow rate.

Users require a ROI for entering the leverage strategy and therefore users will deleverage as the borrow rate approaches the LST interest rate. We believe users need to earn a decent return to justify the leveraged borrow rate exposure and the price impact from swapping when looping. When the incentives of wMATIC borrowing concludes we would like to see the Polygon v3 interest rate updated to introduce a small Base and lower Slope1 parameter such that the interest rate at the Uoptimal point is around 90% of the higher LST yield.

Our comment above is specific to wMATIC. With respect to wETH we believe, in general, the Base value should be retained as it dampens the volatility of the interest rate curve when the borrow rate at the Uoptimal is the same.

There are some good points in this proposal which is lower rates to encourage usage. However, in it is current form we do not support this proposal.

[ChaosLabs]

We agree that the sub-optimal utilization of WETH and WMATIC should be addressed to optimize both protocol utilization and revenue and assure sufficient liquidity for suppliers.

We stand with the observation below:

@MarcZeller

there’s a real potential added value to track reference yield for LSTs (stETH lido yield) & (MaticX yield) and adjust the slope1 parameters algorithmically to a range targeting LST yield. There’s little governance added value to set these parameters and by definition, the DAO will always run behind the market.

Until such a solution is put in place, the main issue is that any decision will always run behind the market, meaning that rates will be too low or too high. @MarcZeller pointed out that setting low borrowing rates at UOptimal could yield unwanted behaviors. Moreover, if we seek to increase utilization around the equilibrium borrowing interest rate, the guideline should be to decrease the sensitivity of the utilization rate to staking yield changes. Since borrowing interest rates are in direct correlation with staking yields, an ideal solution that can handle staking reward variability will decrease the sensitivity of the utilization rate to borrowing interest rate changes. Therefore, we support decreasing the base interest rate to 0 (zero) while maintaining the Variable Slope 1 at 3.8. This way, we will achieve two goals:

1. An immediate increase in utilization to keep the equilibrium interest rate
2. Decreased sensitivity of utilization to interest rate changes and, therefore, to staking yield changes

Illustration of a steep slope (left) and gentle slope (right) showing how the exact interest rate change (dotted red lines) results in a much more significant change in utilization (blue lines) on the gentle slope than on the steep slope.

For MATIC, as utilization has not been stable and given current staking yields, we think setting Slope 1 to 4.3% is a bit low. Furthermore, in light of TokenLogic’s input concerning ongoing and upcoming partnerships, we suggest postponing these updates for the moment. If the community wishes to decrease Variable Slope 1 at this time, we would advise setting it at 4.6%, allowing for a slight margin from the staking APR and the utilization levels observed recently.

[Gauntlet]

Thanks all for the additional thoughts and suggestions for our IR proposal. To reiterate, we aim to encourage more WETH/WMATIC Emode borrowing in order to increase revenue for Aave, as well as to improve the risk profile for Ethereum and Polygon LSTs. Staking yields have been declining, leaving insufficient ROI for users looking to put on recursive borrowing strategies.

We echo @MarcZeller that a well-designed, automated mechanism for slope 1 to track LST yield will bring significant improvement. Designing such a mechanism is necessarily complex, and we give brief thoughts below.

Regarding borrowing behavior for v2 markets, our proposal aims to prevent WMATIC and WETH supply from flowing backward from v3 into v2. Promoting continued migration from v2 to v3 markets can naturally follow up on this proposal.

On setting slope 1 below LST staking yield

We continue to recommend setting slope 1 to 3.3% for WETH and 4.3% for WMATIC. The current 3.8% slope 1 for WETH is below rETH yields and in line with average yields for wstETH through August 2023. The current 6.1% and proposed 4.6% slope 1 for WMATIC are also significantly higher than the current 4.2% yield for stMATIC.

• Enables the most flexibility for all LSTs to successfully participate in recursive borrowing strategies, which will reduce dependency on LST liquidity for liquidations, thereby improving their risk profile.
  • Staking yields vary among LSTs; in particular, rETH had a roughly 3.2% annual yield throughout June-July 2023.
  • LST liquidity has been declining, as noted in our original forum post.
• Provides buffer for profitable recursive strategy in the event LST yields decline, allowing borrow rates to be “ahead” of the market, and also reduces the sensitivity of WETH/WMATIC utilization to staking yield changes.
  • LST yields have been declining over the past few months, as noted in our original forum post. As an example, rETH yield currently sits at ~3.5% APR as of 2023.08.25.
• Finally, given Slope 2 at 80%, no excess risk with the equilibrium utilization > Uopt.
  • Under the proposed change (setting slope 1 to 3.3%), equilibrium utilization will not be materially higher than Uopt (less than 1% greater than Uopt).

protocol	symbol	New Utilization (equilibrium borrow rate = current borrow rate)	New Utilization (equilibrium borrow rate = max historical LST yield, past 90 days)
aave_v2	WETH	0.8005	0.8043
arbitrum_aave_v3	WETH	0.8004	0.8043
ethereum_aave_v3	WETH	0.8001	0.8043
optimism_aave_v3	WETH	0.8004	0.8043
polygon_aave_v3	WETH	0.5303	0.8043
polygon_aave_v2	WMATIC	0.4171	0.7523
polygon_aave_v3	WMATIC	0.7275	0.7523

On stMATIC/MaticX/WMATIC behavior

With the current staking APR for stMATIC at 4.2%, WMATIC borrowing APR at 5.5%, and WMATIC incentives at 0.3% APR, users currently are not seeing ROI for the recursive borrowing strategy, even with the current incentives.

This is reflected in the changing borrow distribution over time against stMATIC collateral. Currently, 55% of borrows supported by stMATIC are not in Emode, compared to 45% in Emode.

Reducing Slope 1 for WMATIC is the most effective method to revitalize WMATIC borrowing demand. We reiterate our recommendation to lower Slope 1 to 4.3% in order to facilitate more WMATIC borrowing against Polygon LSTs. As noted above, this will not only increase revenue from WMATIC borrowing, as calculated in our impact table above, but also improve the risk profile of stMATIC and MaticX by reducing dependency on Polygon LST liquidity for liquidations.

On automated mechanism to enable slope 1 on WETH and WMATIC to track LST yield

Creating a well-designed mechanism to facilitate strong and healthy WETH, WMATIC borrowing is the most natural method to preserve optimal utilization. The challenge here is that insufficiently designed automated targeting mechanisms have their tradeoffs, and can introduce potential attack vectors analogous to JIT liquidity attacks for Uniswap. These attack vectors essentially allow for a strategic user to profit and seize higher yield at the expense of less sophisticated lenders, which over time may negatively impact the protocol. We provide a more in-depth response here.

[maxicrouton]

Hey!
I really like the topic.
But doesn’t this simply transfer income from asset providers to the Aave protocol and borrowers (leveraged stacking)?
When do you plan to publish the snapshot ?

[Gauntlet]

Thank you all for the feedback. After continued analysis over the past couple of weeks, we aim to move forward with the following. Combined together, the following will help balance between risk with insufficient liquidation capacity during large market moves, while continuing to incentivize recursive LST borrowing strategies.

Proposal to lower WETH Uopt on Ethereum v3 from 90% to 80%.

Currently, there may be insufficient WETH to facilitate liquidations during increased volatility - this proposal aims to double the available WETH for liquidations should utilization for WETH be at Uopt = kink.

Our recommendation with more analysis is here. We aim to put up snapshot on 2023.09.11.

---

Snapshot for WETH IR slope 1

We aim to put up a snapshot to adjust WETH Slope 1 on 2023.09.11. It will include the following options.

Options	Asset	Description	Recommendation
1	WETH	(lower slope 1 below staking yield)	lower slope 1 on all v3 to 3.3%
2	WETH	(match v3 Ethereum)	remove base rate of 1% on all v3, excl Ethereum v3

We continue to recommend option 1 in order to facilitate LST recursive strategies due to lowering LST yields.

---

WMATIC next steps

Due to WMATIC utilization moving towards the kink in the past week, we aim to delay our snapshot to move WMATIC Slope 1 to 4.3%. Analysis of WMATIC utilization shows that WMATIC borrows have not increased, but rather WMATIC supply has decreased and converted into STMATIC. These STMATIC positions are collateralizing stablecoin borrowing, which we have analyzed considerably above.

To address the decreasing MATIC LST liquidity as highlighted here and in our original post we aim to put up a snapshot to lower STMATIC and MaticX non-emode LT next week as well, to mitigate risky stablecoin borrowing.

We will provide more analysis in that proposal.

[Gauntlet]

We put up the snapshots for the WETH IR slope 1 updates here and WETH Uopt updates here. Voting begins in 1 day and lasts for 3 days.

[Gauntlet]

Thanks for the community’s participation in the snapshot— option 1, lowering slope 1 of WETH on all v3 to 3.3% had the highest vote (66.6%). However, given the community preferred not to move forward with lowering WETH Uopt, we no longer recommend changing slope 1 to 3.3% on v3 Ethereum due to increased risks of high utilization. We continue to recommend lowering slope 1 of WETH to 3.3% on all v3 markets except Ethereum. We target publishing the AIP on Monday, September 25.

[Gauntlet]

As a followup to the community, option 1 was chosen on Gauntlet’s previous snapshot, which lowered slope 1 for WETH on all v3 to 3.3%. On the other hand, option 2 recommended to remove the base rate for WETH to match v3 Ethereum. The AIP is published here.

Option 1 was intended to be a superset of option 2 and also intended to lower the base rate to 0 for WETH, in order to drive utilization to Uopt. This language was not specified in the snapshot. To be extra clear with the community, we put out this new snapshot specifically for lowering the base rate to 0 for WETH on v3 markets, here.

---

For a more concrete example on how IR will change -

• Currently, utilization on Arbitrum v3 is 43%, Uopt = 80%.

  • The AIP to lower WETH slope 1 to 3.3% on all v3 markets except Ethereum will adjust current Arbitrum WETH borrow rate from ~3% to ~2.7%. We will put up this AIP today.
  • When utilization = Uopt, WETH borrow rate will be 4.3% with the AIP.

• This rate may be too high to drive utilization to Uopt, given current LST staking yields.

  • Removing the Base Rate will cause the WETH borrow rate to be 3.3% at Uopt = 80%, which is more conducive to LST strategies.
  • This may help to increase WETH borrows to drive utilization to Uopt.

[Gauntlet]

As an update, AIP-327 to set WETH slope 1 to 3.3% on v3 markets, excluding Ethereum v3, has been executed.

For the recommendation to lower WETH base rate to 0, we’ve moved forward with Option 1 from the snapshot for V3 Arbitrum and Optimism which have higher LST TVL and expected impact from this proposal relative to other V3 markets. The AIP is published here.

[Gauntlet]

AIP-340 has been executed. We thank the community for their participation.