Implicit conversion from float -43.87755102040816 to int loses precision