The probative value of forensic science evidence (such as a shoeprint) varies widely depending on how the evidence and hypothesis of interest is characterized. This article uses a likelihood ratio (LR) approach to identify the probative value of forensic science evidence. It argues that the “evidence” component should be characterized as a “reported match,” and that the hypothesis component should be characterized as “the matching person or object is the source of the crime scene sample.” This characterization of the LR forces examiners to incorporate risks from sample mixups and examiner error into their match statistics. But will legal decisionmakers be sensitive to the extent to which examiners' statistical testimony accounts for various potential sources of risk and error? A controlled experiment with 315 jury-eligible jurors who received a shoeprint match statistic in a hypothetical burglary case finds that, contrary to normative theory, people are more persuaded by statistical testimony that ignores various error risks than by testimony that is objectively stronger by virtue of taking those risks into account. The experiment also finds that jurors are relatively unresponsive to exposure of those risks by a defense attorney on cross-examination. These results support and extend previous research that finds many people are confused about how to evaluate the risk of error associated with expert forensic testimony.