Risk assessment instruments have been the subject of a number of validation studies which have mainly examined the psychometric properties known primarily from psychological test development (objectivity, reliability and validity). Hardly any attention was paid to the fact that validation of forensic risk assessment instruments is confronted with a whole row of methodical challenges. Risk assessments include a quantitative and a qualitative component in that they state the probability (quantitative) of a particular offense (qualitative) to occur. To disregard the probabilistic nature of risk calculations leads to methodically faulty assumptions on the predictive validity of an instrument and what represents a suitable statistical method to test it. For example, ROC analyses are considered to be state of the art in the validation of risk assessment instruments. This method does however not take into account the probabilistic nature of prognoses and its results can be interpreted only to a limited degree. ROC analyses for example disregard certain aspects of an instrument's calibration which might lead in an instrument's validation to high ROC values while demonstrating only low validity. Further shortcomings of validation studies are that they ignore changes of risk dispositions or that they don't differentiate between offense specific risks (e. g. any recidivism vs. violent or sexual recidivism). The paper discusses and reviews different quality criteria of risk assessment instruments in view of methodological as well as practical issues. Many of these criteria have been ignored so far in the scientific discourse even though they are essential to the evaluation of the validity and the scope of indication of an instrument.