This article examines an educational experiment with a unique combination of 3 elements: homework, the use of information and communication technology and a large degree of freedom of choice (student autonomy). More particularly, we study the effectiveness of a web-based intelligent tutoring system (ITS) that a school offers to its students as optional homework, in a sample of 355 first-year secondary students, using an experimental design. The results show that whether students make this noncompulsory homework in the web-based ITS is dependent on their prior achievement and their teacher, and if they practice, they tend to choose easier modules. Students thus do not seem to optimize learning gains, but rather balance their perceived value of practicing and their "expectancy," as found previously for regular (compulsory, non-ITS) homework. Regarding the value of choice, findings suggest that students may be poor judges of their most beneficial exercises pattern. Differentiation in ITS does not guarantee learning, unless scaffolding is tightly linked to it. In line with previous research, this research shows that practice does matter, though at varying intensities. Yet it also clarifies that motivation to make an effort is required, which adds to the lessons learned from homework research; both the perceived value and expectancy are at play and can be enhanced, for example, by teachers.