Purpose The aim of this study was to examine the predictive validity of the German adaptation of the Static‐2002 and to compare it with the results of the Static‐99. Method The predictive validity of Static‐2002 was investigated in a sample of n = 452 sexual offenders released from the Austrian Prison System. The instrument was coded retrospectively using file information. Afterwards, the predictive estimates of the Static‐2002 were related to officially documented reconviction data. Results The Static‐2002 was found to have large effect sizes for predicting sexual, violent, and general recidivism (area under the receiver operating characteristic curve [AUC] = .78, .75, .75, respectively) for the total offender sample. For the general sexual offender sample, it did not predict reoffence significantly better than the Static‐99 ([AUC] = .77, .72, and .72, respectively). However, the Static‐2002 was found to perform better in the child molesters subsample (AUC = .83, .74, .75, respectively) than in rapists (AUC = .66, .72, .73, respectively), even though these differences did not reach statistical significance. The Static‐2002 was found to have incremental predictive validity beyond the Static‐99, but only for general and not for sexual or violent recidivism. Conclusions Although a replacement of the Static‐99 by the Static‐2002 cannot be recommended, it has to be emphasized that the Static‐2002 yielded large effect sizes in predicting general, violent, and sexual recidivism and, therefore, nevertheless represents an excellent measure for capturing the reoffence risk of sexual offenders.