I would like to add another idea here. For V1-V2 timeline:
It strikes me that it is a little weird that both ETH and USDC on aave are 80/85 max borrow/liquidation.
It seems at first glance like the usdc liquidation risk should be… very low since it has no price volatility.
I believe this is likely bc liquidations in aave involve exchanging sets of tokens rather than just liquidating the collateralizing asset(unlike maker, where everything is liquidated for dai), and so the final return is dependent on the volatility of both assets, not just the collateralizing asset.
This makes me think that basing these ratios on the collateralizing asset is wrong, since it has no direct relationship to how risky a borrower is. Risk parameters should be related to correlations of assets lent and borrowed, higher correlations on each side would mean better collateralization ratios. Ratio trading should give you much better terms and higher leverage than straight levering volatile crypto. I propose we try to fix this issue in a future version of Agave.
One way of getting around actually calculating volatility and correlational data on chain or with oracles is just creating asset categories which we consider highly correlated, and assign some “beta” parameter:
- large cap crypto (>$1 B?) beta ~ 2
- small cap crypto (<$1 B) beta ~ 3 or 4
- LPs beta ~ 1
- stablecoins beta = 0
- low-vol. non-stable assets (paxg?) beta ~ 1
- etc
We could then beta weight the assets on the borrow and the lending side, and take the difference of the 2 sides for determining the riskiness of a borrower, and giving more aggressive/conservative lending terms based on that.
It is possible 1 risk parameter (called here beta) may not be enough to cover all our bases, but this seems like it would give a significant improvement in efficiency of collateralized lending, open up interesting future projects, and even help us get closer to a system that automatically assigns risk parameters.