Introduction

The composability of the DeFi ecosystem implies that risks from an individual component flow into all dependent systems. Liquidity is at the heart of the Unlockd Protocol; it enables the protocol's operations and user experience.

This documentation presents a framework to assess token risk for V2 of the Unlockd Protocol. The risk methodology considers the market, counter-party, and smart contract risks for the tokens selected for the Unlockd Protocol, aiming to contribute to higher risk standards within DeFi.

Unlockd is and will remain to be, a fully data-driven protocol. As such, all parameters fixed within this document, even those which the DAO will set, are subject to future changes if available data suggests a more optimal parameter configuration.

Furthermore, in line with our commitment to continuous improvement, the Unlockd Protocol will be introducing Machine Learning techniques and hyperparameter optimization as additional tools in our risk model. As more data is gathered, these advanced methodologies will be leveraged to further enhance the performance and effectiveness of our risk assessment model.

By combining rigorous risk assessment frameworks, data-driven approaches, and the incorporation of advanced techniques like machine learning, Unlockd aims to provide both borrowers and lenders with a robust and continuously evolving protocol that prioritizes risk management and optimization within the DeFi ecosystem.

Considerations

1. The low boundary of our providers will be taken as the price value for any asset to consider the Median Relative Error and reduce the protocol’s risk.

2. The confidence parameter on the Price Object should not be considered as it displays redundant information already shown in the MRE and is taken into account in the lower boundary within the low attribute.

3. All parameters like ε are subject to further discussion and are not to be treated as definitive. Many of them, which will be fixed in this document, should be voted as Unlockd transitions to a DAO, and further analysis will be done as more data is collected from real usage of the protocol. The error terms are mainly added to ensure robustness and stress testing on the model simulations to prevent any kind of overfitting over the data.

4. For the Standard Deviation calculation, the returns are set to be a finite, normally distributed variable, and we will be using the incomplete array of prices as population $∴ ddof = 1$ for inference purposes.

5. As more data is gathered the protocol risk framework will take into account higher moments (skewness, kurtosis, etc) into the models.