The principal causes of error in the rubber model are discussed and, in some simple cases, evaluated. It is shown that errors due to the configuration of the membrane and ball spin can be reduced without limit by decreasing the scale of deformation of the membrane. Energy loss due to friction is the limiting factor, and a method is described by which this may be measured (as a function of ball speed), results being given for balls of various sizes rolling over a pure rubber surface.In general, a reduction in the area of the membrane and an increase in the ball diameter are advantageous, subject to constructional limitations. For a path length of about one foot and a ball diameter of one inch, errors in ball energy can be maintained below 2%.