Implicit conversion from float -13.25 to int loses precision