This paper presents how the kinematic and potential energy analysis of self-adaptive robotic legs can help to improve their performances with respect to their ability to overcome obstacles and the required actuation torque to do so. Self-adaptive leg mechanisms, inspired by the underactuated linkages used in grasping, generally rely on a single degree of freedom (DOF) to generate a trajectory at its endpoint that is appropriate for walking applications. When colliding with an unexpected obstacle, a second DOF in the leg automatically engages and creates a motion allowing the leg to overcome said obstacle. Since this behavior is obtained mechanically, with no sensor or control, these robotic legs are referred to as self-adaptive. In this paper, the conditions for the passive adaptation to obstacles are first briefly recalled. Then, the range of obstacles for which this adaptation is possible is determined through the analysis, using potential energy, of the mechanism workspace and it is shown how the results are connected to its kinematics. In particular, the influence of the shape of the terminal link of the leg is discussed with two compared examples. Finally, practical designs and especially the relative advantages of various locking mechanisms, required to improve stability during the support phase of the leg trajectory, are discussed.