Predictive models, which includes forecasting models, are used to study all types of conflict and political violence, including civil wars, international conflict, terrorism, genocide, and protests. These models are defined as those where the researcher explicitly values predictive performance when building and analyzing the model. This is different from inferential models, where the researcher values the accurate operationalization of a theory, and experimental or quasi-experimental designs where the focus is on the estimation of a causal effect. Researchers employ preditive models to guide policy, to assess the importance of variables, to test and compare theories, and for the development of research methods. In addition to these practical applications, there are more fundamental arguments, rooted in the philosophy of science, as to why these models should be used to advance conflict research. Their use has led to numerous substantive findings. For example, while inferential models largely support the democratic peace hypothesis, predictive models have shown mixed results and have been used to refine the scope of the argument. Among the more robust findings are the presence of nonlinear relationships and the importance of dependencies in all types of conflict data. These findings have implications for how researchers model conflict processes. As predictive models become more common and more integrated into the study of conflict, it is important that researchers understand their underlying components to use them appropriately.