Collateral Factors: How to Win the Lending Race

[bryancolligan]

Collateral Factors: How to Win the Lending Race

A lot of factors go into what makes a good race car, but the most obvious is speed. Sure traction control and pit security matters, but if the other cars are faster than you, you’ll end up safely in last place. Driving a much slower car is akin to operating a lending protocol with lower collateral factors.

Looping strategies make up about a third of Collateral Deposits on Compound Mainnet. People are using leverage to loop up assets as many times as possible without crashing – reminiscent of race cars speeding around a track.

• Hold your Position on Aave, and your car can go between 100 and 140mph.

  • [10 to 14 loops.]

• Hold your Position on Compound, and your car will only go 50 to 80mph.

  • [5 to 8 loops.]

Collateral Factors are the governor on the engines of these vehicles controlled by dao governance. The only way that Compound can win the race is to hope that Aave can’t finish it. This eventuality would mean that catastrophe hits only Aave markets, and assumes people migrate from Aave to Compound.

Trying to compete in a race where we are not evenly matched is also expensive, because we must compensate for a lack of matched power and handling, with incentives. Even when positions are liquidated at 140mph, Aave profits from the user’s liquidation.

Compound is at a disadvantage with Aave on two points:

1. Collateral Factors for Looped Assets.
2. Supply Side Liquidity.

In order to properly compete with Aave, Compound must run grant incentive campaigns to attract Looping TVL via Contango, Vaultcraft and Other Leverage Products. We could rent a balance sheet for supply side liquidity, so we have capacity, and when utilization gets too high, users can adjust their positions appropriately.

Risk Management could institute an E-Mode for Compound, and match our looping parameters with Aave, but Compound DAO needs to consider if this is a viable strategy. As the Collateral Factor rises, the track gets more narrow, and dangerous. The only way Compound can win is if the Drivers on Aave get over levered and blow up.

Consider Renzo:

Take a look at the Parameters for the Mainnet ETH and wstETH markets below for ezETH, as well as the TVL held within those markets:

ezETH on Aave:

• Collateral Factor is 93.
• Liquidation Factor is 95.
• Aave holds $685M in ezETH.

ezETH on Compound:

• Collateral Factor is 88.
• Liquidation Factor is 91.
• Compound holds $80M in ezETH.

In terms of Collateral Factor, Compound is more conservative than Aave. This means Racers are limited in their max speed (or leverage).

Liquidation Factor means that Compound incentivizes their Liquidators by offering a larger premium on rekt positions – pushing to clear way protocol debt as quickly as possible.

In terms of TVL, Aave holds 8.5x more ezETH than Compound. If Aave has more generous parameters, deeper supply-side liquidity, and provides greater access to yield – how can Compound hope to compete?

Aave’s parameters for most looped assets are more risk-on, and as such, they attract far more deposits.

The Conservative Choice

The Conservative Choice for the individual is Compound, because if you are only going 50 to 80mph, you are more likely to survive a de-correlation event driven by slashing or liquidity constraints.

With our current strategy, we need Aave to implode, in order to win – and Compound DAO needs to consider whether or not this is a viable strategy.

It’s almost as if our parameters are telling our users, “Just Use Aave.”

• x.com

[mexilc]

CFs arent even the most important part of Comets

Feels like a basic misunderstanding of how compound works

[cmrn]

@mexilc

CFs arent even the most important part of Comets

There aren’t unlimited variables to play with. CFs are a strong one imo.

Research like this is indeed valuable. Thanks for it @bryancolligan

[bryancolligan]

Many conversations have been brought by delegates around Utilization and how to increase the utilization of comets. We interviewed different integrations and liquidity managers on why they didn’t utilize Compound as much as AAVE and time and time again it comes back to leverage, the market is requesting more loops and leverage.

As a market participant the protocol needs to expand with the market requests or leave the upside on the table for competitors. If this is what the community chooses to do, the growth team will 100% align. It is our stance that we should surface the market opportunity have a discourse and then align to what the community wants.

Always open to any other suggestions you have.

[misher]

Could Compound offer a higher collateral factor in return for a higher liquidation penalty to balance the risk? With a big message you have to click ok on to access it.

If not, I’d be ok matching aave for the ones that don’t seem crazy. But there should be big red warnings and suggestions not to go above X so people are less likely to blame Comp if things go wrong.

[Sudipan_Chainrisk]

[CHAINRISK] E-MODE EVALUATION FOR COMPOUND

Thanks @bryancolligan for bringing this up on the forum. If the DAO feels that the race of collateral factors is relevant for Compound to win, Chainrisk herein presents an analysis on the implementation of E-MODE.

Collateral factors are a critical component of any lending protocol, determining how much a user can borrow against their deposited assets. AAVE offers higher collateral factors (E-Mode Category) than Compound, allowing users to leverage their assets more aggressively. This difference is fundamental in attracting liquidity and user engagement.

This disparity enables AAVE users to execute more aggressive borrowing strategies, enhancing their capital efficiency and potential returns.

E-MODE

E-MODE (Efficient Mode) in AAVE allows for increased capital efficiency by enabling higher loan-to-value (LTV) ratios based on asset correlations, thus enhancing borrowing power while capping risk exposure.

---

AAVE E-MODE Analysis:

ETH Correlated:

The diagram shows the correlation between ETH and related assets like WETH, osETH and other co-related assets over the past six months (June to December 2024). It highlights how these assets’ prices move in relation to changes in ETH’s price.

Stablecoins:

The diagram depicts the correlation between USDT and USDC in terms of USD over the past six months (June to December 2024). It illustrates how the prices of these stablecoins move in relation to fluctuations in the USD.

LST Correlated:

The diagram illustrates the correlation between WSTETH, ETHX, and RSTETH in terms of ETH over the past six months (June to December 2024). It highlights how the prices of these liquid staking tokens respond to fluctuations in ETH’s price.

E-MODE Borrow Distribution across 6 chains

E-MODE Borrow Percent v/s Total Borrow

The above table shows that a significant portion of their borrowing activity is conducted under this efficient mode.

Aave V3: E-Mode Bad Debt across 6 chains:

Total Bad Debt: $32,318

Given the inherent correlation in price dynamics of assets under E-mode, the incidence of bad debt within the E-mode is comparatively minimized.

---

Analyzing Compound V3

The following table summarizes the total borrowing and collateral across various chains in the Compound v3:

---

Comparison of Aave V3 and Compound V3

---

Looping Strategies

Looping strategies are essential for maximizing asset utilization within both protocols. On Compound Mainnet, looping strategies account for approximately one-third of total collateral deposits. Users leverage these strategies to maximize their asset utilization while managing liquidation risks effectively.

Looping Example

Initial Setup:

• Deposit: 1 osETH
• User Borrow at Max LTV / Collateral Factor
• Swap borrowed ETH for osETH again Loop.

AAVE

Max LTV: 93%

COMPOUND

Collateral Factor: 80%

Aave allows for greater total net exposure due to its higher Max LTV of 93%, resulting in a total net exposure of approximately 7.77 ETH after the 10th loop, compared to 4.57 ETH on Compound with an Collateral Factor of 80%.

While both protocols allow for looping, the higher collateral factor in AAVE enables users to achieve greater leverage and potentially higher returns.

---

Implementation

We suggest establishing separate markets specifically tailored for e-mode assets. This would allow all co-related assets to be grouped within a distinct market environment, effectively isolating risks associated with these correlated assets. Exchange rate oracles can be integrated to enhance capital efficiency by enabling borrowers to operate with reduced collateral requirements, thereby optimizing capital utilization. However, reliance on dynamic exchange rates can increase the protocol’s exposure to insolvency risks, particularly in scenarios involving liquid staking token (LST) or base asset depegging.

If the community shows sufficient interest, we’re happy to help in implementation of E-MODE on Compound.

[bryancolligan]

A quick read on this suggest higher Collateral Factors such as E-mode accounts for about half of the utilization for AAVE on ETH Mainnet, Arbitrum, OP and BASE chain. Is this the net-net conclusion?

[Sudipan_Chainrisk]

Yes, it’s one of the main conclusions we drew from our analysis. Based on our analysis, E-MODE borrowing represents a substantial proportion of the aggregate borrowing on Aave.

[mexilc]

Stop copying Aave (emode) and Morpho (sandbox). Come up with something new or innovative.

Copy paste wont make Compound a leader in antything

[bryancolligan]

thanks for the Feedback. We need higher collateral factors to be able to compete.
whether we call it e-mode or extra-risky-collateral or any other branding name the intention is the same higher CF == higher utilization.

[misher]

Agreed, I’m fine with high as long as there’s a warning for users. Buyer beware after all. As for the point about copying Aave, I think it’s fine to copy Aave to stay competitive while also innovating.

[Gauntlet]

Gauntlet appreciates the initiated discussion and here we would like to add quantitative considerations around Collateral Factors. Gauntlet recommends that modifying Collateral factors (CFs) without robust fallback measures can expose the protocol as well as the users to larger risk surface area. This is also why we recommend Compound to maintain a buffer between the Liquidation Penalty (LP) and the Liquidation Factor (LF), ensuring an extra margin of safety against rapid price drops or price-feed anomalies. This is specifically true as seen during the market events of Jan 2024 and Aug 2024.

1. Absence of a kill-switch

Aave’s success with higher CFs (or LTVs) partially hinges on its ability to trigger an on-chain emergency shutdown (kill-switch) if extreme volatility (caused due to slashing, etc) or manipulation arises. Compound, however, has no comparable feature. While governance can pause or upgrade markets, these processes are manual and less efficient in terms of response time, offering less agility in crisis scenarios. Without a rapid-response mechanism, drastically raising CFs could leave Compound more vulnerable to liquidation spirals, oracle exploits, or market shocks.

Gauntlet recommends that the pause functionality should be implemented as new “kill-switch” functions in smart contracts similar to Aave, and/or added as assertions directly in the smart contract business logic that can pause certain market functions based on the deviation between LSD and market prices.

2. Lack of Correlated-Asset Price Oracles

Aave relies on specialized oracles that track price movements among correlated assets (such as ETH and various LSDs) which counteract the possibility of sudden, artificial spikes in the staking rate of a correlated asset. The Correlated-Asset Price Oracles helps prevent scenarios where a compromised oracle feed could artificially inflate the exchange rate. Such an event could otherwise lead to a large accumulation of bad debt. By contrast, Compound’s standard oracle system evaluates each asset independently, lacking mechanisms to detect when two assets lose correlation. This shortcoming makes broad-based high collateral factor settings inherently riskier.

The below example runs through a scenario of upward depeg caused by artificial volatility in staking rate.

Upward Depeg by 10%

1. Initial Setup:

   • User supplies $1,000,000 worth of ezETH ($3,500 per ezETH).
   • Collateral Factor (CF) for ezETH is 88%, allowing a maximum borrow amount of $880,000 or 220 wstETH ($4000 per wstETH) at an exchange rate of ~1.142.

2. wstETH Exchange Rate is artificially increased by 10%:

   • User borrows $880,000 or 220 wstETH when the exchange rate is 1.142.
   • At 1.256 WETH/wstETH, the value of the borrowed amount increases to $967,120.

3. Absorption and Liquidation:

   • User is absorbed due to the increased liability.
   • Collateral (ezETH) is taken by the protocol.
   • With a Liquidation Penalty (LP) of 6%, the user is owed $1,000,000 * (1 - 0.06) = $960,000 or 240 wstETH.
   • Store Front Price Factor (SFPF) is 97%, so the value of collateral sold is $1,000,000 * 0.97 = $970,000.
   • Since the collateral is sold when wstETH is 1.256 (~$4400 per wstETH), the protocol receives $970,000 / 1.256 = 220.45 wstETH tokens.

4. Repeg and Protocol Losses:

   • Upon repeg, the 220.45 tokens are now worth $881,800.

5. Shortfall Calculation:

   • Liability: $967,120.
   • Value from collateral sale: $881,800 or 220.45 wstETH.
   • Total shortfall per user: $967,120 - $881,800 = $85,320.

6. Multiple Users Example:

   • If 10 users each exploit this:
     • Total borrowed wstETH: 10 * $967,120 = $9,671,200.
     • Total collateral value after liquidation: 10 * $881,800 = $8,818,000.
     • Total shortfall: $9,671,200 - $8,818,000 = $853,200

Given the above scenario, a Correlated Oracle Price framework would cap the inflated staking APRs while the kill-switch would trigger an on-chain pause across various functions on the protocol.

3. “Soft E-Mode” Already Exists in Compound V3 (Comet)

Compound V3’s WETH Comet is effectively a borrow-only market, where users can supply a small set of correlated or high-quality assets as collateral to borrow WETH. This structure is already a “soft” E-Mode, but more restrictive than Aave’s multi-asset approach. A full-scale E-Mode in Compound would require allowing the base asset (WETH) to be used both as collateral and the borrowable asset—necessitating sweeping architectural changes.

Furthermore, adopting Aave’s E-Mode is not only incompatible with Compound’s single-borrowable-asset model but could also manifest more risk to the protocol considering the present liquidation mechanism. Implementing many-to-many lending would fragment reserves with the Comet’s inability to pay the borrower back due to a high degree of fragmentation.

4. Rationale for lower CFs and LFs for osETH

It’s important to note that osETH relies on a market rate oracle, which necessitates deliberately lower Collateral Factors (CFs) and Liquidation Factors (LFs) to account for the inherent volatility between osETH and WETH. Consequently, this design makes leveraging osETH capital-inefficient, as the lower CFs and LFs impose tighter constraints to mitigate volatility risks.

Recommendation

While Aave’s E-Mode is effective for its particular ecosystem—enabling higher leverage through correlated-asset oracles and kill-switch safeguards—Compound V3 has neither the mechanism to detect artificial deviations from staking rates nor the rapid-response tools necessary for such an aggressive risk posture. If the community wishes to seek higher CFs, we recommend that the protocol implement a 1. Implementing an on-chain kill-switch and 2. Developing robust correlated-asset oracles

We believe these adjustments provide enhanced protection for the protocol to enable more aggressive CFs while avoiding the disruptive changes and additional risks associated with implementing e-mode. Given the absence of these protective mechanisms, Gauntlet is not comfortable increasing collateral factors higher in these markets at the current market conditions. However, if the community still wishes to enable higher Collateral Factors we suggest implementing a snapshot vote to gauge preference.

[misher]

An on chain kill switch seems like a good idea if it uses a multi-sig with a % of delegates required to enable it?