In our study (N=2189), we divided participants into 6 badge conditions: 1) Role model badges (e.g., Einstein), 2) Personal interest badges (e.g., Movies), 3) Achievement badges (e.g., "Code King"), 4) Choice, 5) Choice with badges always visible, and 6) No badges. Participants played a CS programming game, then used an editor to create their own level. Badges promoted avatar identification (personal interest, role model), player experience (achievement, role model), intrinsic motivation (achievement, role model), and self-efficacy (role model) during both the game and the editor. Independent of badges, avatar identification promoted player experience, intrinsic motivation, and self-efficacy. Additionally, avatar identification promoted greater overall time spent in both the game and the editor, and led to significantly higher overall quality of the completed game levels (as rated by 3 independent externally trained QA testers). Our study has implications for the design of badge systems and sheds new light on the effects of avatar identification on play and making.