Objective To analyze the effect of virtual reality combined with game-based task-oriented training （TOT） on balance function and fall efficacy in stroke patients.Methods A total of 114 stroke rehabilitation patients treated from January 2024 to April 2025 were selected and divided into a conventional group (n=57), undergoing conventional balance training, and a research group (n=57), undergoing virtual reality combined with game-based task-oriented training, both trained for 6 weeks.The Berg Balance Scale (BBS) and the Modified Falls Efficacy Scale (MFES) were used to evaluate the patients′ balance function and fall efficacy before training, at 4 weeks and 6 weeks of training.The Gait Assessment part of the Tinetti Scale was used to assess the gait function of the patients before training and at 6 weeks of training, and the Timed Up and Go Test (TUGT) was performed, and the fall incidence rate within 6 weeks of intervention in the two groups was analyzed.Results A total of 55 cases in the routine group and 54 cases in the study group completed the study.The time effect, between-group effect, and interaction effect of the balance function and fall efficacy scores were significant (all P＜0.05).After 6 weeks of training, the Tinetti gait score of the research group was higher than that of the conventional group, the TUGT time was shorter than that of the conventional group, and the fall incidence rate within 6 weeks of intervention in the research group was significantly lower than that of the conventional group (all P＜0.05).Conclusion Virtual reality combined with game-based task-oriented training, through immersive scene construction, phased game-based tasks, and real-time interactive feedback design training strategies, can enhance patients′ balance function and fall efficacy, improve gait function, and thus reduce the risk of falls from the dual pathways of self-identity reshaping and precise intervention.