Sorry, you need to enable JavaScript to visit this website.
Skip directly to content

Results on Pricing American Options with Reinforcement Learning

35.00 mins
Dr. Daniel Bloch
Thu 21 Sep 2023

Some of the main difficulties when pricing an American put option with reinforcement learning is to make sure that the continuation value one step before maturity is equal to the European put option, and that the Snell envelope is equal to the strike at maturity and decreases exponentially with time-to-maturity. We chose to assess the performance of existing RL networks based on those two criteria and propose some modifications to improve results. We benchmark our model with the continuous BAW on non-dividend paying stock and obtain a satisfactory outcome.