We explore the effectiveness of repeating the student-teacher match on test scores, for the universe of 8th graders in Chile using information on all student-teacher matches across multiple subjects and years, and a national, anonymous measure of test scores. Using a fixed-effect and a regression discontinuity approach, we find that repeating matches has a robust positive effect on test scores. We show that this positive effect aggregates up to the student, class, and school-level, and also has longer-term effects on university admission exams. As channels, we find a significant positive effect on attendance, progression, student behaviour and teacher expectations. Reallocating teachers to classes with which they are familiar appears to offer a feasible strategy to improve student performance at low cost.