Implicit conversion from float -30.665322580645167 to int loses precision