If an empirically based classification system could be developed and made operational, those offenders requiring physical confinement would be so confined, only those requiring imprisonment would be in such a setting, and those requiring little or no supervison would be free or nearly free of all physical and psychological restraints. In addition to being empirically based, a good classification system focusing on risk determination must be highly reliable, easy to use, and correspond as much as possible to commonsense considerations. To the extent possible, many aspects of the predictive system should rely on routinely collected, highly objective data such as age, number or arrests, and age at first arrest. In evaluating classification systems, the crude but salient measure of recidivism should be used as the primary criterion. The classification system operating in California incorporates several elements that are basic to an ideal system. One aspect of the classification/screening effort is the presentence diagnostic study provided by the corrections department to the court. Another procedure that would fit into the framework of screening for risk is the 1-year parole discharge. This procedure is based on the finding that those who remain arrest-free for the first year of parole have about a 90 percent chance of completing the second year without serious legal involvements. In attempting to implement a new classification system along these lines, resistance to change must be anticipated. Eight references are included.