This study examined temporal interrelations among math self-concept and math self-efficacy and their differential relations to two commonly applied math achievement indicators, that is, math achievement test scores and school grades in math. It based on a longitudinal design with four annual assessments of self-concept, self-efficacy, school grades, and achievement test scores in math during secondary school (Years 6 to 9, N = 3,209 German students). Prior math self-concept was positively related to subsequent math self-efficacy, but prior math self-efficacy was unrelated to subsequent math self-concept. Considering math grades, reciprocal relations could be found for math self-concept. Math self-efficacy and math grades showed a unidirectional relation with former math grades being related to later math self-efficacy. Considering math test scores, reciprocal relations could be found for math self-efficacy. The relation between math test scores and math self-concept was unidirectional, former math test scores being related to later math self-concept. The relations among self-concept, self-efficacy, and the two types of achievement indicators in math did not vary across time waves which covered adolescent years. Gender invariance was demonstrated with regard to the interrelations of math self-concept and math self-efficacy as well as for their relations to both math achievement indicators.