This study investigates the potential of using a newly developed sub-game in the Magical Garden research platform as a method of teaching preschool children subitizing. Thirty preschool children played the game, identifying formation with four, five and six objects. Some formations with five objects were scaffolded by changing the sprites of some of the objects. Results showed that children were significantly faster and accurate at formations with four objects than five and six. No significant difference was found between five and six objects, which were also considerably similar in both time and accuracy. Analysis suggests that perceptual subitizing was used to a greater degree when presented with four objects and that counting was used for higher numerosities. The study showed that there were some problems with the study design, with formations being more difficult at formations with five objects. Suggestions are made for how the game should be altered, including a dynamic difficulty changing component to account for the large individual differences. Additionally, it is suggested to lower the amount of time the formations are shown in order to elicit the use of conceptual subitizing and focus on lowering the difficulty in terms of the number of items and formation pattern rather than altering time. There was no significant difference for performance when comparing scaffolded formations with non-scaffolded formations. The reason could be that the formations were not subitized. Future studies should include the suggested changes and conduct longitudinal studies looking at improvements over time and whether children are improving.