The Dispute Outcome Expectations score is a probabilistic measure of how a hypothetical militarized dispute between two states would end, taking into account each side's raw military capabilities.
The DOE scores are the output of an ensemble of machine learning models. We fit the models on the Militarized Interstate Disputes data and then use them to calculate predicted probabilities for every dyad–year from 1816 to 2012.
If you use the DOE scores, please cite the associated American Journal of Political Science article.
DOE scores are available for download from Dataverse. The current version of the dataset is 2.0.
The scores come in two flavors: directed and undirected. The directed scores give the probability of the initiator winning, the target winning, or a stalemate. The undirected scores give the probability of each state winning (or a stalemate), assuming a 50-50 chance of either side being the initiator.
The code used to create the DOE scores is available on GitHub.
For a full description of the methods and data used to create the DOE scores, see our American Journal of Political Science article "Prediction, Proxies, and Power".
You may also be interested in the working paper version.
The Dispute Outcome Expectations project was created by Rob Carroll and Brenton Kenkel.