• The research addresses the lack of robust risk measurement frameworks for the rapidly growing tokenised RWA market, which exceeded tens of billions of dollars by 2025.
• It argues that standard Value-at-Risk (VaR) models understate tail losses in these markets due to episodic trading volumes and abrupt liquidity shifts.
• The study uses an XGBoost gradient boosting algorithm to forecast volatility, incorporating features like volume impulse indicators and cross-asset volatility spreads.
• Empirical testing was conducted on a portfolio of PAX Gold (PAXG) and Ondo Finance (ONDO) using data from April 2024 to April 2026.
• The findings demonstrate that the proposed L-VaR framework provides superior tail loss coverage compared to standard models, especially during low-liquidity regimes.

‍