There has been significant TVL moving into crvUSD LP pools on Convex and Curve. We present an analysis of potential opportunities and risks for Gro DAO in whitelisting crvUSD/USDT and crvUSD/USDC pools on Convex as new strategies.
crvUSD is an algormithic asset-backed stablecoin. Its key innovation is in drastically reducing losses during liquidation through the use of Lending-Liquidating AMM Algorithm (LLAMMA). LLAMMA spreads liquidity into bands and introduces “devaluation” of partial collateral when collateral price drops, and recovers the collateral when price bounces back; In a historical back-test, if the price of the collateral drops and then rebounds, the liquidation loss under LLAMMA was as little as 1% upon liquidation, if the collateral loan is spread across a range of 20%. This is a significant saving compared to standard liquidation penalty charges between 5%-10%, if the collateral price went back.
LLAMMA works by creating “bands”, where collateral is deposited into a range of them. Similar to ticks in Uniswap-V3, each band has a non-linear price relationship to each other. Health factor of a position is then computed by calculating the effective collateral value based on user’s share on each of the band. As the price of collateral drops, the deposited collateral in the top of the band ranges would be swapped into crvUSD gradually (devaluation), while the health factor could remain positive (healthy).
The rebalancing process of selling and buying back collateral assets within the pool is conducted by external traders or arbitrageurs.The AMM under LLAMMA was specifically designed so that changes in external price data (P_ORACLE) would lead to larger deviations in the price in the LLAMMA Pool (P_AMM) to incentivize arbitrageurs to rebalance the pool and perform the partial liquidations or de-liquidations of the pool’s assets.
Blockworks newsletter described the peg mechanics and current incentive behaviors well, emphasis ours in bold. TLDR is that mechanics encourage repegging, including overcollateralization, burning crvUSD mechanic, and borrow rate changes.
To maintain the crvUSD peg, Curve utilizes dynamic borrow rates and Pegkeepers. The latter works by minting and burning crvUSD in its AMM pools. More tokens are minted when crvUSD trades above its peg, and the inverse applies when crvUSD is below the peg. The dynamic borrowing rate is calculated based on how far the crvUSD price is from its peg and the amount of Pegkeeper debt in the system. A decrease in the crvUSD price or Pegkeeper debt outstanding would result in an increase in the borrowing rate, and vice versa. This is because if the debt is close to zero, the protocol can’t put a lot of upwards pressure on the crvUSD price by removing supply if needed. When the crvUSD price decreases, a higher borrowing rate forces stablecoin holders to redeem tokens. In other words, borrowing rates increase to encourage debt reduction and decrease to encourage borrowing. All in all, the mechanism seems to have worked relatively well. A ~0.5% depeg isn’t bad at all, considering the circumstances.
Nevertheless, it’s worth noting that the crvUSD supply was only slightly below 100M tokens when the exploit happened, meaning that crvUSD hasn’t yet proven its stability as a widely adopted stablecoin. Furthermore, before the hack, the difference between the crvUSD supply and the amount borrowed was 20M tokens, but the figures quickly converged on August 1. In other words, the Pegkeepers ran out of tokens to burn and could therefore not affect the price by restricting supply. The price could have dropped further if there was more sell pressure, but the peg floor would have probably still been quite high since crvUSD is currently overcollateralized by almost 1.5x, meaning that arbitrageurs would have swiftly stepped in.
Time since launch: 3 months
TVL: $126.05m
CRV Liquidity: $21.22m
Audit: audit report