htho:
By knowing wich balls are near a ball, and where the bowls wall is, i can determine the positions of all balls.
That sounds like a leap of faith to me. I don't think you have very much control over the range of the RFID, so it may not be easy to reliably sense all the touching balls without also sensing some that are further away. Assuming you can manage that, then calculating the physical geometry of a set of balls based on which ones are touching is probably not feasible unless you constrain the balls to a known layout (in a line, for example) - if they're just tumbling around in a bucket I don't think you have enough information to calculate their positions.