This experimental study investigated the state (short-term) effects of action video game (AVG) training on arithmetic performance and their persistence over time. In addition, it examined group differences between experienced and novice AVGers. Twenty-nine college students without a prior AVG experience were randomly assigned to one of the two training groups: AVG and non-AVG. After 40 minutes of video game training, the arithmetic problem-solving speed and accuracy of non-AVG group increased, while the AVG group’s arithmetic performance decreased, thus suggesting a possibility of state effects of a non-AVG training on arithmetic performance. The state effects did not persist over time; on a delayed posttest, both groups’ arithmetic performance was similar to their pretraining scores. In addition, there were nonsignificant differences in arithmetic performance between experienced and novice AVGers. Implications for investigating the game mechanics and transfer mechanism between the game and transfer task are discussed.