Pré-publication, Document de travail: The paper develops an Early Warning System (EWS) to identify the build up of vulnerabilities in the external sector of 31 Emerging Markets (EMs) across the period 1995-2017 and avoid the painful sudden reversal of capital flows associated to them. It contributes to the literature on the prediction of financial discontinuities in three ways. First, it uses a discrete choice model to calculate and compare the marginal effect of different domestic and global factors on the probability of a sudden stop materializing. Second, it analyzes the performance of the model with a recursive framework that reflects accurately the information set available to policymakers at the time of the prediction. Third, it investigates the relationship between ex-ante probability of a sudden stop and the ensuing output loss. We find that domestic and global factors contribute to the reversal of capital flows in a comparable way. Our model calls half of the pre-crisis periods, exhibiting a high specificity and a proper timing. Moreover, we find a positive link between the ex-ante probability of a sudden stop and the associated ex-post loss. These results call for an active use of Early Warnings in the policy-making sphere.