Collaboration among students is fundamental for knowledge building and competency development. Nonetheless, the effectiveness of student collaboration depends on the extent that these interactions take place under conditions that favor commitment, trust, and decision-making among those who interact. The sanitary situation and the transition to remote teaching has added new challenges for collaboration given that students interactions are mediated by Information and Communication Technologies (ICTs). In this study we explore the effectiveness of different collaborative relationships on physics and mathematics, from a sample of secondary students from two schools located in rural and urban areas in southern Chile. We used Social Network Analysis (SNA) to map students friendships relations, academic prestige, and collaboration on both courses. Later we combined the collaboration network with friendship and academic prestige on the course, to separate strong from weak friendship working ties, and those among students who enjoy or not academic prestige. Multiple linear regression models showed, on average, positive effects of collaboration on grades. Yet, when isolating the effects of the types of collaboration, the positive effects are observed only between those who display more strong friendship ties. Also, we found differences on the social networks and their effects over grades between both courses, presumably due to their pedagogical nature. With these results we contribute to the literature of collaboration and its effectiveness based on the nature of students relationships, and advocate for the importance of instructional design in fostering appropriate motivations and guidelines for constructive collaboration in the classroom.