Assessed understanding of decision making about repeated uncertain events, using hypothetical and in-vivo prediction tasks modeled after those used in probabilitiy-learning research. Previous studies have been unclear on whether suboptimal choices reflected reasoning errors and lack of strategic thinking, or confounding factors. Quantitative and qualitative analyses of subjects' choices and strategy explanation showed that up to 50% of college students did not fully understand the relative value of different strategies. Only 5 % of subjects preferred a "true" probability matching strategy, included on an in-vivo task. High-school students showed greater misunderstandings. Large gender differences in prediction strategies and in related computational skills were observed. Understanding was discussed in terms of subjects' inferences from knowledge of independence of events, (lack of) computational skills, and correlates of the quantitative nature of prediction tasks. Implications for research on decision making, including need to address individual differences, are discussed.