In this paper, we present the first meta-analysis of the efficacy of game-based interventions on reducing students' levels of maths anxiety. After searching for randomised studies describing game-based interventions to reduce maths anxiety, 22 effect sizes with 913 participants described in 15 peer-review articles met the selection criteria. A random effects meta-analysis indicated a reduction of maths anxiety with a small effect size (mean effect size ES = −0.24, CI = [ − 0.47, −0.01]), marginally significant at 0.05 level but not robust to a leave-one-out sensitivity analysis. Several factors moderated the results: non-digital games were more effective, while digital games had a negligible mean effect size of ES = −0.10, CI = [ − 0.24, 0.03]. The effect size was also moderated by the total duration of the intervention, to the advantage of longer interventions, and by the type of gameplay: games had a greater effect on maths anxiety reduction when they promoted collaborative and social interactions. Such features were mainly present in non-digital games, while all bar one of the digital interventions used single-player games. The results obtained, which were particularly weak for digital games, indicated the need to develop and test games explicitly designed for maths-anxious students to increase the impact of game-based interventions. This will require investigation into the relationship between game features and maths anxiety through analysis of the behaviour of anxious and non-anxious students at play. Among the features that an anxiety-aware game could employ, we suggest collaborative gameplay, social interactions, adaptability, features promoting intrinsic motivation and embedding real-time measurements of maths anxiety in the game.