JP Morgan quants are building deep hedging 2.0

New model uses Bellman technique to learn general derivatives hedging strategies

JP Morgan quants are working on the next iteration of the firm’s machine learning hedging engine – a version that learns to hedge any book of options rather than just one book at a time.

The bank’s so-called deep hedging engine uses machine learning to work out how to hedge derivatives books from raw data, factoring in real-world market frictions such as transaction costs. Quants have described it as the most exciting research in derivatives pricing and risk management.

But the current engine