Improper priors can cause problems in comparing models using Bayes factors, but a new method has been developed to address this issue. By controlling how quickly prior certainty changes, well-defined Bayes factors can be computed even with divergent priors. It has been shown that exceptions to the previous issues exist, with certain improper priors leading to reliable Bayes factors. The Shrinkage prior by Stein is one such example. A simple solution has been proposed to deal with any resulting problems. A Monte Carlo experiment has confirmed the effectiveness of this new approach.