Comparing strong and weak models by fitting them to computer-generated data