This paper describes three methods of measuring the coefficient of restitution (CoR) for two different types of ball-on-ball collision. The first collision type (for which two different CoR measurement procedures are described) is a static, hanging steel ball forming part of a Newton's cradle arrangement, which is then hit by its adjacent identical ball, swinging down from an angle. The second scenario (for which one CoR measurement procedure is described) is a snooker ball interaction in which the cue ball rolls in, and collides with, a static coloured ball (both balls being of a resin composition). The investigation requires only readily available and inexpensive equipment, together with an open-source video analysis programme, called "Tracker." Perhaps surprisingly, the experiment yields widely differing CoR values for the two types of interaction. This variance cannot be solely accounted for by the difference in physical properties of the respective balls' compositions. The paper then describes, in theoretical terms, the details of the dynamic interactions in each example, and hence validates the surprising discrepancy in the two results obtained empirically.