Implicit conversion from float -4.08163265306122 to int loses precision