The effectiveness of a 10-session social problem-solving training (SPST) was evaluated in two independent studies in a juvenile justice (JJ) setting. In both studies, we aimed to examine main intervention effects on social problem-solving skills and recidivism, as well as differential effects as modulated by anger, anger regulation, and single nucleotide polymorphisms in the COMT gene. In Study 1, a randomized controlled trial (RCT), 289 male detainees (M_age= 14.95 years) were randomly assigned to SPST or treatment-as-usual (TAU). In Study 2, a pre-post community implementation, 187 youth (M_age = 16.03 years) on probation were assessed before and after SPST. No significant main effects of SPST on social problem solving or recidivism were shown in either study. With regard to differential effects, among youth in detention,COMT haplotypes predicted intervention effects on state anger. Moreover, independent of SPST, inward anger expression was associated with an increase in state anger from pre- to post and an increase in state anger with a decrease in social problem solving. Among youth on probation, COMT haplotypes predicted social problem-solving skills, and, in turn, an increase in social problem-solving skills decreased the odds of recidivism after SPST. The lack of main effects of SPST may be due to low program integrity in JJ settings. Juveniles' emotional and genetic characteristics might modulate the effectiveness of interventions in JJ settings. We recommend studying large samples to substantiate this observation.