Objective: This study tested whether simulator-based training of pilot responses to unexpected or novel events can be improved by including unpredictability and variability in training scenarios. Background: Current regulations allow for highly predictable and invariable training, which may not be sufficient to prepare pilots for unexpected or novel situations in-flight. Training for surprise will become mandatory in the near future. Method: Using an aircraft model largely unfamiliar to the participants, one group of 10 pilots (the unpredictable and variable [U/V] group) practiced responses to controllability issues in a relatively U/V manner. A control group of another 10 pilots practiced the same failures in a highly predictable and invariable manner. After the practice, performance of all pilots was tested in a surprise scenario, in which the pilots had to apply the learned knowledge. To control for surprise habituation and familiarization with the controls, two control tests were included. Results: Whereas the U/V group required more time than the control group to identify failures during the practice, the results indicated superior understanding and performance in the U/V group as compared to the control group in the surprise test. There were no significant differences between the groups in surprise or performance in the control tests. Conclusion: Given the results, we conclude that organizing pilot training in a more U/V way improves transfer of training to unexpected situations in-flight. Application: The outcomes suggest that the inclusion of U/V simulator training scenarios is important when training pilots for unexpected situations.