There is a rich literature highlighting that pathogens are generally better adapted to infect local than novel hosts, and a separate seemingly contradictory literature indicating that novel pathogens pose the greatest threat to biodiversity and public health. Here, using Batrachochytrium dendrobatidis, the fungus associated with worldwide amphibian declines, we test the hypothesis that there is enough variance in novel host-pathogen outcomes to pose substantial risk of pathogen introductions despite local adaptation being common. Our continental-scale, common garden experiment and global-scale meta-analysis demonstrate that local amphibian-fungal interactions result in higher pathogen prevalence, pathogen growth, and host mortality, but novel interactions led to strikingly variable consequences with the greatest risk occurring when susceptible hosts and virulent strains interacted. Thus, while most pathogen introductions are benign, enough variance exists in novel host-pathogen outcomes that moving organisms around the planet greatly increases the chance of pathogen introductions causing profound harm.