We examined the effects of high- versus low-anxiety conditions during video-based training of anticipation judgments using international-level badminton players facing serves and the transfer to high-anxiety and field-based conditions. Players were assigned to a high-anxiety training (HA), low-anxiety training (LA) or control group (CON) in a pretraining-posttest design. In the pre- and posttest, players anticipated serves from video and on court under high- and low-anxiety conditions. In the video-based high-anxiety pretest, anticipation response accuracy was lower and final fixations shorter when compared with the low-anxiety pretest. In the low-anxiety posttest, HA and LA demonstrated greater accuracy of judgments and longer final fixations compared with pretest and CON. In the high-anxiety posttest, HA maintained accuracy when compared with the low-anxiety posttest, whereas LA had lower accuracy. In the on-court posttest, the training groups demonstrated greater accuracy of judgments compared with the pretest and CON.