# Methodology for handling short american options in a back test

Given that an American option can be exercised at any time, how does one handle algorithmically shorting an American option in a back test? I am not sure what the best practice is to simulate the early exercise for selling American Options in a backtest.

There are two approaches I can think of:

1. Assume that the option gets exercised if it crosses an arbitrary percent in the money.
2. Treat the option the same way you would handle a European option.

An American put without dividends is exercised when $S$ hits an optimal exercise boundary (e.g. if $S=0$ one would always exercise because the maximum payoff is reached). Finding the optimal exercise boundary for an AM put is however still an open question, some researches assume an explicit functional and fit it to the model.