Child development research on overconfidence suggests that the bias is present and persistent in preschoolers and kindergartners. However, little is known about what drives overconfidence among young decision-makers, how it changes over a large number of repetitions, and whether such changes differ by gender or age. The current experimental study analyzes data from 60 children, aged 4 years 0 months to 6 years 10 months, who played 60 turns of the Children's Gambling Task and provided regular estimates on their performance. A video intervention, designed to demonstrate the consequences of disadvantageous choices, was tested in a double-blind randomized controlled trial to assess its impact on overconfidence. The results show that every third participant remained overconfident even after 60 trials and constant feedback. Unlike previously reported, gender seems to be a determining factor in this process. Lastly, providing additional information through a video intervention appears to have no impact on participants' overconfidence levels.