Implicit conversion from float -98.08988764044943 to int loses precision