The importance of self-related constructs in predicting academic achievement has been increasingly emphasized in recent decades. Typically, bivariate associations of self-related variables with achievements have been reported. Research quantifying the combined predictive power of more than two self-variables has been scarce. Moreover, except for the academic self-concept, these variables have almost always been measured across domains, i.e., without considering the specifics of individual school subjects. The current study aimed to statistically predict academic achievement (operationalized via school grades) in three major subjects (Chinese (native language), mathematics, and English (foreign language)) by using subject-tied scales, namely academic self-concept, conscientiousness, need for cognition, perseverance of effort, and consistency of interest. The sample comprised 791 Chinese adolescents. Each scale was related separately to each of the three school subjects. Hierarchical linear regression analyses were run. The control variable, biological sex, accounted for 2% of Chinese grades and 8% of English grades, but not of mathematics grades. Adding subject-specific self-concept scales increased the explained variance to 7% (Chinese), 16% (mathematics), and 32% (English). Further additions to the other four self-related scales did not increase the variances that were accounted for. The discussion underlines the relevance of subject-specific academic self-concepts as predictors for subject-tied academic achievements.