Implicit conversion from float -18.897637795275585 to int loses precision