The estimation of loss distributions for dynamic portfolios requires the simulation of scenarios represent- ing realistic joint dynamics of their components, with particular importance devoted to the simulation of tail risk scenarios. Commonly used parametric models have been successful in applications involving a small number of assets, but may not be scalable to large or heterogeneous portfolios involving multiple asset classes.
We propose a novel data-driven approach for the simulation of realistic multi-asset scenarios with a particular focus on the accurate estimation of tail risk for a given class of static and dynamic portfolios selected by the user. By exploiting the joint elicitability property of Value-at-Risk (VaR) and Expected Shortfall (ES), we design a Generative Adversarial Network (GAN) architecture capable of learning to simulate price scenarios that preserve tail risk features for these benchmark trading strategies, leading to consistent estimators for their Value-at-Risk and Expected Shortfall. We demonstrate the accuracy and scalability of our method via extensive simulation experiments using synthetic and market data. Our results show that, in contrast to other data-driven scenario generators, our proposed scenario simulation method correctly captures tail risk for both static and dynamic portfolios.