Thank you for your replys.
We understand that there are different speeds depending on which part of the rotating mass you measure. But if the curve were designed to contact the ball in "only" two points (and not half ball or something like that) ? Having the benefit of center the ball and also another point of contact to transfer the energy to the ball.
Something like this (very quick design to think in that option):
http://imgur.com/a/Y0e4F