The Sexual Violence Risk–20 (SVR-20) is the most commonly used structured professional judgment guideline for sexual offender risk assessment and risk management planning. Using a prospective longitudinal research design the authors evaluated the predictive accuracy of the SVR-20, of its subscales, and of the individual items in relation to different recidivism criteria and offender subgroups by analyzing a total sample of 493 sexual offenders released from the Austrian Prison System. Findings indicate good predictive accuracy for the prediction of sexual recidivism for the total sample (area under the receiver operating characteristic curve [AUC] = .72) as well as for the rapist subgroup (n = 221, AUC = .71) and the child molester subsample (n = 249, AUC = .77). However, the results also revealed some inconsistency in the predictive power of the SVR-20 depending on recidivism criterion and offender subgroup. Furthermore, it can be pointed out that the subscales and individual items substantially differ in their association with future sexual and nonsexual violence.