Background: Researchers debate whether game-based learning (GBL) affects students' achievement emotions, as past studies yielded mixed results. Objectives: This study determines the overall effect of GBL on students' positive achievement emotion and their negative achievement emotions (PAEs, NAEs), along with their moderators. Methods: This meta-analysis examined true or quasi-experimental studies, reported in English or Chinese, and with students (ranging from primary schools to universities) who did not report physiological and/or psychological illnesses. Results: Compared to traditional instruction, GBL had a positive effect on students' PAEs (g =0.526, k = 39, 95% CI = [0.319, 0.733]) and a negative effect on NAEs (g = -0.517, k = 22, 95% CI = [-0.709, -0.324]). GBL's positive effects on students' PAEs were (a) largest in middle school and otherwise generally larger in higher grades, and (b) largest for competition GBLs (compared to cooperative or individual GBLs). GBL's negative effects on students' NAEs were (a) larger in collectivist cultures than individualistic ones, (b) progressively larger in primary school, middle school, and university, (c) larger for non-digital games than digital games, (d) smallest for individual games, (e) progressively larger in mathematics, social sciences, language, science, and highest in engineering and technology, and (f) smallest for extremely short or extremely long intervention durations (very short: up to 45 min; very long: over 2 months). Take Away: GBL often increases students' PAEs and decreases their NAEs, though their effect sizes differ across cultural values, grade level, game type, academic subject, or intervention duration. Hence, educators should consider these aspects when designing/selecting education games for their students and instructional contexts.