Who Fails and Why: Student Trajectories and Early Prediction of Performance in an Introductory Programming Course

This study examines failure in introductory programming courses, commonly known as CS1, in Chilean higher education by combining academic trajectory analysis with early-risk prediction models. We analyzed a cohort of 994 students from a Chilean technical university enrolled during the first academic semester of 2025, with a 46% failure rate, integrating pre-university academic and admission variables (e.g., mathematics and language indicators, as well as baseline diagnostic measures when available), sociodemographic information, and within-semester performance indicators. Group differences were assessed using non-parametric tests, and predictive performance was evaluated under two realistic information-availability scenarios: (i) pre-university variables only and (ii) variables available up to the first major written examination (C1). The results show statistically significant differences between students who passed and those who failed, with indicators of quantitative preparedness and, most notably, C1 performance emerging as the strongest signals of risk. In the pre-university scenario, models achieved acceptable discrimination (AUC ≈ 0.77), whereas incorporating C1 substantially improved discriminative performance (AUC ≈ 0.92) and increased precision in identifying at-risk students while reducing false positives. These findings support a staged institutional strategy: broad, low-cost preventive support before the semester begins, followed by more targeted and intensive interventions after C1, thereby enabling more efficient early-warning systems in high-stakes first-year courses.