Implicit conversion from float -83.87755102040816 to int loses precision