Deep Hedging to Manage Tail Risk

TL;DR

This paper introduces a deep learning-based approach for tail risk hedging in portfolios, effectively reducing extreme losses while considering market frictions and constraints.

Contribution

It extends the Deep Hedging framework by using neural networks to optimize convex risk measures like CVaR and ES for tail risk management.

Findings

Achieves significant 99% CVaR reduction in simulations.

Demonstrates robustness under market frictions and constraints.

Provides practical insights into strategy adaptation.

Abstract

Extending Buehler et al.'s 2019 Deep Hedging paradigm, we innovatively employ deep neural networks to parameterize convex-risk minimization (CVaR/ES) for the portfolio tail-risk hedging problem. Through comprehensive numerical experiments on crisis-era bootstrap market simulators -- customizable with transaction costs, risk budgets, liquidity constraints, and market impact -- our end-to-end framework not only achieves significant one-day 99% CVaR reduction but also yields practical insights into friction-aware strategy adaptation, demonstrating robustness and operational viability in realistic markets.