Problem-solving skills in creative, open-ended domains are both important and little understood. These domains are generally ill-structured, have extremely large exploration spaces, and require high levels of specialized skill in order to produce quality solutions. We investigate problem-solving behavior in one such domain, the scientific-discovery game "Foldit". Our goal is to discover differentiating patterns and understand what distinguishes high and low levels of problem-solving skill. To address the challenges posed by the scale, complexity, and ill-structuredness of "Foldit" solver behavior data, we devise an iterative visualization-based methodology and use this methodology to design a concise, meaning-rich visualization of the problem-solving process in "Foldit". We use this visualization to identify key patterns in problem-solving approaches, and report how these patterns distinguish high-performing solvers in this domain. [For the full proceedings, see ED596512.]