It is argued that researchers must first undo the existing sentencing regime in order to allow a meaningful analysis of the potential of risk assessment based on observable characteristics and the associated use of statistical treatment rules in sentencing. The largely unrecognized conceptual difficulties with the current risk assessment literature imply that there is a need to start at the beginning in order to assess the potential benefits of a statistical sentencing policy. It is seen as time to abandon old conceptual frameworks and begin work on research which pays explicit attention to the current treatment regime. This will allow those to directly inform current policy and make valid inferences about the predictability of high risk behavior. Over several decades, criminologists have devoted their efforts to the problem of identifying high-rate offenders. These efforts have focused primarily on criminal history records as predictors of future behavior. These efforts to identify high-risk offenders are viewed as ineffective. This paper illustrates the fundamental problem with identifying high rate offenders so as to subject them to harsher punishments via a statistical treatment rule. References