In the last two decades, researchers have conducted several empirical evaluations, involving thousands of subjects, to understand the use of models in software development. The results of these evaluations show that most of the subjects make informal use of the models, which is known as ‘modeling as sketch’. In this paper, we present an experiment that compares UML-based and DSL-based modeling when subjects model a part of a commercial video game. In the comparison, we have used objective and subjective measures, in contrast to other works that focus either on objective measures to evaluate modeling performance or on subjective measures to analyze modeling styles. Our results reveal that subjects underestimate the potential of their own models. Our finding is relevant for the design of future evaluations and for the teaching and adoption of modeling. If users correctly assess their models, they might leverage their potential as programs.