Outcome prediction in alcoholism therapy is of major sociopolitical and economic significance. Instruments based on psychotherapeutic processes are lacking. Therefore, treatment processes of 64 chronic alcohol dependent patients have been investigated at three time-points, t(1) (week 3), t(2) (month 6), and t(3) (month 12) during the first year of a comprehensive outpatient treatment program, guaranteeing strictly controlled alcohol abstinence. Main focus of the study was the prediction of cumulative abstinence probability over a follow-up period of up to 4 years based on these treatment processes. One hundred and seventy-five video recordings of therapy sessions were analyzed with the behavior observational system VAMP (Video-Assisted Monitoring of Psychotherapeutic Processes in Chronic Psychiatric Disease). Patients' self-rating of treatment processes was measured with questionnaires for self-efficacy, abstinence confidence, self-consciousness and stress coping. Prediction of cumulative abstinence probability was determined with Cox regression analysis. By integrating the observer rated process variables with the highest predictive validity, the composite score TOPPS (Therapy Orientation by Process Prediction Score) was constructed. It includes the process variables experience of resources, abstinence self-efficacy, implicit craving, relapse alertness, relapse risk, disease concept, dysfunctional therapeutic engagement, and dysfunctional problem solving of current problems. Whereas patients' self-rating of treatment processes was insufficiently predictive, the TOPPS strongly predicted four-year abstinence probability at any of the 3 time-points (p<0.001). The results suggest to validate the item combination described in the TOPPS in addiction therapy as a treatment guideline of individual relapse prevention strategies.