We introduce the new, substantially updated, and revised version of the Dyadic Militarized Interstate Disputes (MIDs) dataset. We discuss the underlying logic of constructing dyadic MIDs and demonstrate that these operations generate significant differences between the actual occurrence and properties of MID dyads and those extracted from machine-generated programs such as EUGene, or from the MID participant dataset. We provide some descriptive measures of dyadic MIDs over the period of 1816 to 2010 and compare some of the key dyadic results on the correlates of MIDs using different datasets. We discuss the theoretical and empirical implications of our results.