Implicit conversion from float -21.980198019801975 to int loses precision