Design considerations

The ARCx Credit protocol has been designed to avoid excessive exposure to any single party, and to rely on empirical data and rational incentives instead of trust.

  • The DeFi Credit Score is based on real statistical indicators of credit risk, and contains no subjective analysis of a debtor’s profile based on their identity. Trusting a brand name fund, trading desk or other market participant has proven to be extremely hazardous, subject to significant tail risks and the possibility of fraud. We believe that trusting on-chain data levels the playing field for market participants and will prove itself as a more reliable indicator for credit risk.

  • The rules of the DeFi Credit Score are transparent and easy to understand. Instead of building a “black box” machine learning model that ingests hundreds of data points to return a result, we enable lenders to do their own research to understand the counter-party risk of our users. Through explaining the rules clearly to both parties and publishing updated scores on-chain each day, we are providing the tools for lenders to evaluate the performance of our risk modeling more than other credit scores are willing or able to do. Ultimately the market determines the probative value of our scoring, and will price their liquidity accordingly.

  • ARCx Credit and DeFi Credit Score system health KPIs are monitored and publicly available to track for lenders, borrowers and investors alike. We also manage a number of unique control parameters that influence profitability. Making this data and the process by which different parameters are updated more transparent is critical to building trust with market participants.

  • The three-tiered vault design and the addition of a dynamic credit limit based on previous borrowing actively prevent exploitation and debt concentration risk for the protocol. If a user deposits $1 worth of ETH, and borrows perfectly until they have a Score of 999, they will not then be able to borrow a large sum of money in Vault C.

Last updated