The present invention relates to methods for determining whether a subject having prostate cancer is at an increased risk for relapse or rapid relapse. It is based, at least in part, on the results of a comprehensive genome analysis of 273 prostate cancer samples, which indicate that the percentage of large size CNVs predicts prostate cancer relapse. In certain embodiments, a method for determining whether a prostate cancer patient has an increased risk of suffering a relapse or a rapid relapse comprises determining the number and size of CNVs in a sample and determining a large size ratio, where if the large size ratio exceeds a particular threshold, the patient is deemed to be at an increased risk for relapse or rapid relapse.