A year ago, a group of people without any chemistry background played the game Foldit and resolved a protein problem that stumped scientists for over 12 years. They did it in ten days. This talk focuses on the most powerful question that arose from our Foldit discoveries: is possible to achieve general world-class mastery in a specific domain through voluntary prolonged game play? I will discuss our recent work that applies the same principles to early-math education with focus on fractions, proportional reasoning and algebra. Specifically, we turn the education process into a data driven science that, through use of student and teacher interactions discovers best interventions for each unique student as well as the best teacher practices. Our engaged learning process optimizes not just for student mastery of mathematics, but also for their long term engagement and self-identification with STEM material in general. I will describe our approach to achieving both engagement and mastery in our classroom trials across US. The talk will conclude with our recent studies towards creating a teacher community around these tools that allows for rapid teacher development as well as student improvement.