Sentencing judges and parole release authorities are increasingly using statistical risk assessments to guide their decision-making. Risk assessment instruments rely on statistical research and modeling to predict an individual’s chance of recidivating based on information about the individual like age and number of prior arrests. These instruments are subject to serious statistical and legal limitations and these limitations are exacerbated by imperfect use in practice. This Note analyzes the statistical, legal, and practical weaknesses of risk assessment and argues that although risk assessment is a valuable criminal justice tool, its use must be subject to careful oversight. Because the procedural and constitutional protection available to defendants at sentencing is substantially more robust than the protection available to inmates at parole, this Note advocates, in accord with a recent draft of the Model Penal Code, that statistical risk assessment should be used only at sentencing and not at parole. This Note further suggests potential modifications to the Model Penal Code to more carefully balance the potential pitfalls and benefits of the use of risk assessment at sentencing.