Implicit conversion from float -50.308988764044926 to int loses precision