The ability to predict the future is key to intelligent behavior and good governance. It is also famously difficult, even for the most expert individuals in any domain. Fortunately, collective intelligence can greatly help push the forecasting frontier, especially in domains where too many variables are involved for individual experts, and too few structured data are available to feed machine-learning models. Here we present the results of a large-scale crowd forecasting experiment conducted jointly by Hypermind and the Johns Hopkins Center for Health Security to predict infectious disease outbreaks, such as Ebola, Measles, or coronavirus disease 2019 (COVID-19). While most experts predicted poorly, the crowd outperformed even the most accurate individual. We explain why and how crowds can offer wise predictions given the right combination of expertise, diversity, independence, and aggregation methods.