The present study investigates how the different modes of game-design triggers learning outcomes, focusing on peer learning and intergroup competition. A problem-solving science game was developed to help secondary students to learn about the motion of objects. Participants (N = 110) from an urban middle school were randomly assigned to four game-design conditions (individual-competition, individual-no-competition, peer-competition, and peer-no-competition). The results indicated that the peer-competition and peer-no-competition groups outperformed those in the individual-competition and individual-no-competition groups in terms of conceptual knowledge. Additionally, peer-competition groups exhibited higher interest and value and lower tension than those in the individual gameplay groups. Patterns of learning behavior revealed the emergence of in situ science-related problem solving in the peer-competition mode of GBL. Implications on the effectiveness of game-design for GBL are discussed.