I'm sure the fact that this would be Morgan's last game was somewhere out there on the Internet, and certainly should have been part of the training data, in which case I would guess that a more nuanced prompt to the AI could have elicited a more robust output.