Approve API3 Data Feeds for use within Compound Finance

[ugurmersin]

Hey @Gauntlet team!

we really appreciate the feedback!
I’ll go through your feedback point by point!

@Gauntlet

1. Bridging to API3’s OEV Network:
   • Since searchers/liquidators must bridge to API3’s OEV Network, it introduces different dynamics compared to Mainnet which employs use of private mempools, potentially limiting the network effects and competition found on Mainnet with added switching costs to operate multiple instances of liquidation mechanisms.

This is entirely correct. Searchers need to “onboard” by moving funds to OEV Network and doing so requires more effort than to simply liquidate without any prerequesites. This is mainly due to the fact that we require collateral to disincentivize dishonest behavior and act as spam protection on OEV Network. However, if onboarding will give you a competitive edge against searchers that aren’t, you’re guaranteed to outperform them. We’ve already seen this happen on some protocols on Blast, where searchers were wondering why they aren’t able to perform any liquidations anymore, and all that took was one that did onboard.

@Gauntlet

2. Flashloan Restrictions:
   • The use of flash loans is not entirely possible or is met with friction, as searchers must front a percentage of collateral and bridge it to the chain. This collateral is locked when a bid is submitted, adding a layer of complexity to the process. It remains unlocked until a bid is awarded.
   • The absence of inefficiency of flashloans will likely reduce capital efficiency and limit the types of participants able to engage in liquidations. This could lead to a less diverse and potentially less robust liquidator ecosystem and may cause insolvencies to the protocol.

This isn’t quite right. Flashloans aren’t restricted at all on the target chain. The only thing that is required is that sufficient funds to make a bid are available on OEV Network, but how much is needed there is also solely dependent on a) what the collateral factor is set at and b) the bid amount (which is solely dependent on the competition on OEV Network). For point a) the collateral factor is currently set at 10% but this isn’t definitive and in testing. It was set at 20% of the bid amount prior but was already halved after searcher feedback and realization that it was set to conservatively.
As for point b): If the barriers of participation are too high for OEV Network, that will also result in significantly lower bids and as such significantly lower collateral requirements. While this will result in less value being returned to Compound as bidding isn’t as competitive as it could be, it has 0 effects on the actual ability to perform liquidations. I can win a bid that is worth 0.000001 ETH and still liquidate a million dollar position if nobody else is bidding for the same update. As such, people that take on the capital will more than likely outperform those who don’t.

@Gauntlet

3. Latency and Liquidation Timing:
   • The proposed maximum latency of 60 seconds plus bridging time to Mainnet is significantly longer than current liquidation speeds on Compound. This increased latency introduces additional risk considering compound’s liquidation mechanisms, especially during periods of high market volatility.

The latency only exists for the base feed. Data from the OEV Network is real-time and the only latency that exists is how long it takes the searcher to bring the data onto the target chain to perform the respective liquidations. Compound is already exposed to similar latency as the currently utilized Chainlink price feeds are subject to the Chainlink networks ability to push the real-time data onto the respective chain after reaching consensus amongs each other, which is not instantanious and akin to the “real-world” price. The only thing certain about data feeds is that they are outdated (if even by a second) compared to the “real-world”.

From experience, searchers are much more sophisticated actors than most node operators and i trust their ability to identify opportunities and publish prices on-chain much more than node operators that aren’t as thoroughly incentivized (a.k.a it makes no difference if the price is 1-2 seconds too late on your typical oracle (fyi: this already regularly happens as the oracles aren’t “max bidding” when gas spikes) whereas searchers make loads of money the faster they are).

@Gauntlet

4. Collateral Requirements and Risk Management:
   • From the documentation it appears that 5% is the minimum collateral requirement to be awarded bids. This collateral requirement introduces capped upside for liquidators, as the potential profit is limited by the collateral amount. Moreover, liquidators face additional directional risk, especially in volatile markets, which could lead to losses even on successful liquidations. This risk exposure may deter participation, potentially reducing overall market efficiency.
   • In essence, this 5% collateral requirement transforms liquidations from a relatively risk-free arbitrage opportunity into a more complex trading strategy that requires careful risk management and market timing. It aligns liquidator incentives more closely with the overall health of the protocol but at the cost of potentially reduced liquidation efficiency.

It’s important to note that it is 5% of the bid they place. In essence, if there is a liquidation opportunity for a position of $1 Million with a liquidation penalty of 10%, the available reward is $100,000. If a searcher bids $90,000 for the price update that gives them the rights to the liquidation, the collateral amount required on OEV Network is 5% of that bid, so $4,500.
Also important to note that this collateral only gets locked when a bid is won and gets released immediately after proving that the won oracle update has been performed. As such searchers can keep reusing the same collateral as long as they behave honestly.

Yes, you are correct that liquidations become riskier since you’re not guaranteed the liquidation, but only the price update that you win, but this risk will more than likely be incorporated into the bidding behavior of searchers on OEV Network (e.g. lead to smaller bid amounts).

---

Just to summerize:

• data is as fast as you (the searcher) want it on the target chain. OEV Network awards real-time data. The only limitation of speed is how quickly it can be made available on the target chain, which regular oracle infrastructure is also already subjected to
• liquidations become slightly more risky to perform, but those who take on the slight increase in risk are guaranteed to always win. If no other searcher is participating on opportunities the searcher doing so can even win bidding minimally. We can already guarantee that this will not happen as a) we’re working closely together with many searcher teams and b) nobody leaves free money on the table.
• none of this is stopping searchers that use the “regular” liquidation route (that is risk free) to continue operating. It is just pretty much guaranteed that no opportunities will ever make their way to them.

Happy to answer any other questions you might have or go into more detail.