Previous research suggests that action video game players (AVGPs) often outperform non-action video game players (NAVGPs) in cognitive tasks. This study compared the precision of visual short-term memory (VSTM) for motion direction between AVGPs and age- and gender-matched NAVGPs. Participants memorized the direction of random dot kinematograms (RDKs) presented sequentially (one to four per trial) and reproduced the direction of a probed RDK after either a short (0.5 s) or long (3 s) delay. Initial training ensured that all participants reached a predefined performance level with a single stimulus, with AVGPs requiring fewer training blocks to meet this criterion. While no significant group differences emerged at short delays, AVGPs showed significantly higher raw precision than NAVGPs in long-delay trials involving a single stimulus. However, this group difference did not reach significance in the corresponding precision parameter estimated by the Standard Mixture Model. To investigate memory-encoding strategies, we applied the resource-rational model (RRM), which formalizes the trade-off between behavioral accuracy and neural cost. Model estimates showed that NAVGPs placed greater weight on neural cost relative to behavioral benefits during encoding, particularly in long-delay trials, leading to reduced precision. In contrast, AVGPs allocated memory resources more efficiently, maintaining higher precision over extended intervals. These findings suggest that AVGPs adopt more effective encoding strategies, dynamically adjusting resource allocation to task demands. This study highlights the utility of resource-rational modeling for understanding cognitive performance differences linked to action video game experience. Future research could further explore how these strategies translate across different cognitive domains.