Background There has been a lot of research on risk factors for recidivism among juvenile offenders, in general, and on individual risk factors, but less focus on subgroups of serious juvenile offenders and prediction of recidivism within these. Objective To find an optimal classification of risk items and to test the predictive value of the resultant factors with respect to severity of recidivism among serious juvenile offenders. Method Seventy static and dynamic risk factors in 1154 juvenile offenders were registered with the Juvenile Forensic Profile. Recidivism data were collected on 728 of these offenders with a time at risk of at least 2 years. After factor analysis, independent sample t-tests were used to indicate differences between recidivists and non-recidivists. Logistic multiple linear regression analyses were used to test the potential predictive value of the factors for violent or serious recidivism. Results A nine-factor solution best accounted for the data. The factors were: antisocial behaviour during treatment, sexual problems, family problems, axis-1 psychopathology, offence characteristics, conscience and empathy, intellectual and social capacities, social network, and substance abuse. Regression analysis showed that the factors antisocial behaviour during treatment, family problems and axis-1 psychopathology were associated with seriousness of recidivism. Conclusions and implications for practice The significance of family problems and antisocial behaviour during treatments suggest that specific attention to these factors may be important in reducing recidivism. The fact that antisocial behaviour during treatment consists mainly of dynamic risk factors is hopeful as these can be influenced by treatment. Consideration of young offenders by subgroup rather than as a homogenous population is likely to yield the best information about risk of serious re-offending and the management of that risk. Copyright (C) 2010 John Wiley & Sons, Ltd.