Learning an Exercise Policy for American Options from Real Data

We study approaches to learning an exercise policy for American options directly from real data. We investigate an approximate policy iteration method, namely, least squares policy iteration (LSPI), for the problem of pricing American options. We also extend the standard least squares Monte Carlo (LSM) method of Longstaff and Schwartz, by composing sample paths from real data. We test the performance of LSPI and LSM on both real and synthetic data. The results show that the exercise policies discovered by LSPI gain larger payoffs than those discovered by LSM, on both real and synthetic data. Furthermore, for both LSPI and LSM, policies discovered based on sample paths composed directly from real data gain larger payoffs than the policies discovered based on sample paths generated by simulation models with the model parameters estimated from real data.