Clearly you need to divide by whatever the range - the maximum count - of the ADC is.
Haven't researched just how the ADC in the ATmega works.
And of course, the only reason ever to use floating point maths, is to provide a human-readable display. If you are not going to display it, you do not need floating point.