Posted by: Nunchi Quant Team
Abstract:
The Nunchi protocol’s current v1 cross-margining system pools a user’s collateral to support all open positions, a significant improvement over isolated margin. However, it does not natively recognize risk-reducing portfolio structures. This post outlines our research into a v2 upgrade that shifts to a true portfolio-based margin methodology, conceptually similar to institutional Value-at-Risk (VaR) frameworks, and details the core challenges we are actively working to solve.
1. The v2 Proposal: From Asset-Level to Portfolio-Level Risk
The proposed v2 system would analyze the risk of a user’s entire portfolio as a single, coherent unit. Its core function is to recognize and quantify the statistical correlations between different assets and instruments.
-
Example: A user provides a Uniswap v3 ETH/USDC LP token as collateral while opening a short position in our IL-Perpetual.
-
v1 System: The LP token is valued as collateral, and a separate initial margin is required for the short IL-Perp.
-
Proposed v2 System: The risk engine would consult an internal covariance matrix, recognize the strong negative correlation (ρ) between these two positions, and calculate the margin requirement on the substantially lower net risk.
-
2. Active Research Areas for Implementation
Transitioning this model to a secure, on-chain reality presents several challenges that form the basis of our current quantitative research. We are actively seeking input on these areas.
-
(A) Dynamic Correlation Modeling: Our primary focus is developing a robust methodology for modeling and maintaining the correlation parameters between our novel yield primitives and their underlyings. We are analyzing historical data and employing factor models to produce a covariance matrix that is both accurate and resilient to manipulation.
-
(B) Tail Risk & Stress Testing: Standard correlation metrics are known to break down during extreme market events. A significant portion of our research is dedicated to stress-testing the model. We are investigating non-parametric approaches and dynamic correlation models that adjust during high volatility to ensure system solvency during “black swan” events.
-
(C) On-Chain Feasibility & Performance: A major engineering challenge is implementing this model in a computationally feasible way on-chain. We are exploring techniques like off-chain computation with on-chain verification and dimensionality reduction to deliver real-time margin calculations without compromising security.
We believe a portfolio-based risk engine is the inevitable next step for on-chain finance and welcome community feedback on these research directions.