Robot-assisted stroke rehabilitation has become popular as one approach to helping patients recover function post-stroke. Robotic rehabilitation requires four important elements to match the robot to the patient: realistic biomechanical robotic elements, an assistive control scheme enabled through the human-robot interface, a task oriented rehabilitation program based on the principles of plasticity, and objective assessment tools to monitor change. This paper reports on a randomized clinical trial utilizing a complete robot-assisted rehabilitation system for the recovery of upper limb function in patients post-stroke. In this study, a seven degree-of-freedom (DOF) upper limb exoskeleton robot (UL-EXO7) is applied in a rehabilitation clinical trial for patients stable post-stroke (greater than six months). Patients had a Fugl-Meyer Score between 16-39, were mentally alert (> 19 on the VA Mini Mental Status Exam) and were between 27 and 70 years of age. Patients were randomly assigned to three groups: bilateral robotic training, unilateral robotic training, and usual care. This study is concerned with the changes in kinematics in the two robotic groups. Both patient groups played eight therapeutic video games over 12 sessions (90 min, two times a week). In each session, patients intensively played the different combination of video games that directly interacted with UL-EXO7 under the supervision of research assistant. At each session, all of the joint angle data was recorded for the evaluation of therapeutic effects. A new assessment metric is reported along with conventional metrics. The experimental result shows that both groups of patients showed consistent improvement with respect to the proposed and conventional metrics.