The literature on entrepreneurial experimentationcritically assumes a positive relationship between persistence and performance.However, this assumption has not been adequately tested using unbiased samplesthat include entrepreneurs who try but never succeed, as well as those whoexperience differing levels of success. We test theories about the relationshipbetween persistence and performance using two large, unbiased samples fromKickstarter and Google Play. We observe a positive relationship between priorsuccess and persistence, but not between persistence and the probability offuture success. We argue it is likely that the positive relationship betweenpersistence and performance observed in prior work is, in fact, the spuriousproduct of a change in the distribution of skills among the persistent, ratherthan learning. Our findings thus raise important questions about the merits ofexperimental approaches to entrepreneurship.