Building a crvUSD Soft-Liquidation Bot

Curve's crvUSD uses a Lending-Liquidating AMM Algorithm (LLAMMA) that replaces discrete, binary liquidations with a continuous process of 'soft-liquidations.' When a borrower's collateral price falls, their position is not immediately liquidated at a single price point. Instead, their collateral is algorithmically converted to crvUSD across a series of price bands, effectively dollar-cost averaging the exit. This mechanism creates a persistent, on-chain arbitrage opportunity: the LLAMMA offers the depreciating collateral at a discount relative to the external market price to incentivize arbitrageurs to absorb it, protecting the protocol from bad debt.

For searchers and MEV teams, this is a technical challenge of monitoring, simulation, and atomic execution. A bot must continuously monitor the health of all active crvUSD loans by reading each user's deposited collateral range from the Controller and the current price from the LLAMMA's internal oracle. The bot calculates the profitability of calling the liquidate function, which allows it to withdraw a portion of the user's devalued collateral by repaying an equivalent amount of crvUSD. The core of the opportunity lies in the price discrepancy: the bot repays crvUSD at the LLAMMA's internal, discounted rate and can immediately sell the acquired collateral for a profit on an external market, such as a Curve stableswap or Tricrypto pool.

The most sophisticated implementations bundle the liquidation and the subsequent sale into a single atomic transaction using flashloans or multi-hop swaps. A bot can borrow crvUSD from a PegKeeper pool, call liquidate to acquire the discounted collateral, swap that collateral back to crvUSD in a liquid Curve pool, and repay the loan, all within one block. The profitability of this operation is highly sensitive to gas costs, the depth of the collateral's liquidity pool, and the speed of the bot relative to competitors. A poorly designed bot risks being front-run or failing to cover its gas costs, while a well-architected one performs a critical keeper function that directly contributes to the solvency and stability of the crvUSD protocol.

Building a reliable soft-liquidation bot requires deep integration with the LLAMMA's get_p and bands state, the Controller's loan data, and the atomic composability of the Curve ecosystem. Teams undertaking this must model the precise gas budget for each liquidation path, implement robust fallback logic for periods of high network congestion or low liquidity, and continuously simulate the profitability of candidate positions. Chainscore Labs can review bot logic for profitability under various market conditions, audit the atomic execution path for security and economic soundness, and help model the impact of bot competition on expected returns.

The core of a crvUSD soft-liquidation bot is the continuous monitoring of borrower health, which is fundamentally different from the discrete health factor used in protocols like Aave. Curve's LLAMMA does not have a single liquidation point. Instead, a user's collateral is deposited across a continuous range of bands, and a soft-liquidation occurs automatically when the price of the collateral moves through these bands. The bot's primary signal is the calculation of a position's health, which represents the proportion of the user's collateral that has not yet been converted into crvUSD by the LLAMMA's internal AMM.

The health of a crvUSD loan is calculated by calling the health_calculator method on the Controller contract, which internally queries the LLAMMA. The calculation sums the user's collateral across all active bands and compares it to their total debt. A health value of 1.0 indicates the position is fully collateralized and no soft-liquidation has occurred. As the collateral price drops and the LLAMMA automatically trades the user's collateral for crvUSD, the health metric decreases. A health value approaching 0.0 signals that the user's collateral has been almost entirely converted, and the position is nearing a hard liquidation state where remaining collateral can be seized. For a bot operator, this metric is the primary trigger for action.

The arbitrage opportunity for a soft-liquidation bot arises from the price discrepancy between the LLAMMA's internal bands and external markets. When the LLAMMA is actively converting collateral to crvUSD, it does so at a price that may be slightly worse than the external market price, creating a spread. A bot can atomically call the liquidate function on the Controller, which repays a portion of the user's debt and receives the user's collateral at the LLAMMA's internal band price. The bot then immediately sells this collateral on an external venue like the crvUSD/USDC pool or a PegKeeper pool to capture the spread. The profitability of this operation is a direct function of the user's health, the depth of the LLAMMA bands, and external market liquidity. Chainscore can review a bot's health monitoring logic and arbitrage execution model to ensure it correctly identifies profitable opportunities without exposing the operator to adverse selection or gas-inefficient execution.