The purpose of our research was to design a card game in order to improve kindergarten children’s arithmetic performance, particularly the dissociation between perceptual variables and numerosity. Unselected children were subdivided into three groups: a control group (CON), a group receiving digital training (DOI) and a group playing with our card game (BAT). To start with, all children were submitted to the ANOVA test to confirm they had similar levels of competence. We then used the nonparametric chi-squared test to compare their performance on three sets of tests : a pre-test (before training), a post-test 1 (immediate) and a post-test 2 (delayed). Results to the chi-squared test show no significant difference between performance in the control group and in the trained groups. The BAT group however achieves significantly better in the enumeration test compared to the CON and DOI groups. To conclude, our card game does enhance enumeration skills but not really the numerosity-perception interference. A larger scale study is now needed to confirm our results.