Previous meta-analyses of the effects of peer tutoring on academic achievement have been plagued with theoretical and methodological flaws. Specifically, these studies have not adopted both fixed and mixed effects models for analyzing the effect size; they have not evaluated the moderating effect of some commonly used parameters, such as comparing same-age reciprocal peer tutoring, same-age nonreciprocal, or cross-age peer tutoring; considered the educational level of tutee or tutor; or properly addressed publication bias. Most studies are confined to specific populations and particular subjects (mainly mathematics and reading), and some studies are confounded by other types of intervention (such as cooperative learning or adult-led tutoring). Hence, there is a compelling need for an updated, comprehensive meta-analysis evaluating the effect of peer tutoring on academic achievement that incorporates advances in methodology, is not confounded by other modes of peer learning, and engages a wide range of participants and various subjects. The present study demonstrates that peer tutoring has a positive impact on academic achievement. The moderators and crucial determinants of the effectiveness of peer tutoring are identified and compared. Moreover, program parameters based on the concepts of role theory and interdependent group contingencies are evaluated. Finally, a preliminary empirical model of the crucial determinants of best practices for peer tutoring on academic achievement is proposed.