The article explores how well we can predict which school track students will enroll in based on various factors. The researchers used a new method called recursive partitioning from machine learning to create predictive models. They analyzed data from a study that followed 2000 students in Switzerland. The results show that while we can predict some school track enrollments, there are still many aspects that we cannot explain. Most errors in predictions happen between similar school tracks, with fewer mistakes in predicting the enrollment of students in the most prestigious and least prestigious tracks.