Classification models predictions are generally not really probabilities, especially when the training data is imbalanced. This leads into issues comparing individuals predictions. This talk is about how to transform these pseudo probabilities into true usable probabilities.