Was my thought too, you're introducing a number of unknown variables that could skew the distribution.