A large body of research has demonstrated a big-fish--little-pond effect (BFLPE) by showing that equally able students have lower academic self-concepts in high-ability schools than in low-ability schools. Although the BFLPE generalizes across many countries, it varies significantly between countries. The reasons for this variation are still largely unclear. This study investigated how explicit school-level tracking (formal division of students into different school types by achievement) and implicit school-level tracking (informal division of students into schools by social background, controlling for school selectivity) were related to the size of the BFLPE in a sample of 41 countries. BFLPE estimates are based on subject-specific mathematics self-concept as assessed in the 2003 Programme for International Student Assessment (PISA) and reported by Seaton, Marsh, and Craven (2009). Results show that the BFLPE was far more pronounced in countries with earlier explicit school-level tracking whereas implicit school-level tracking did not affect the BFLPE. Surprisingly, the strong relationship between the duration of explicit school-level tracking and the BFLPE was "not" mediated by the size of between-school achievement variance (BSAV) although BSAV was strongly associated with both types of tracking. Moreover, results based on the Trends in International Mathematics and Science Study (TIMSS) 2007 show that the BFLPE is elevated already at 4th grade in early selection countries (i.e., before actual selection). The strong relationship between the duration of explicit school-level tracking and the BFLPE was not evident when the BFLPE was estimated by more general self-concept measures as in PISA 2000 and PISA 2006.