Implicit conversion from float -13.896103896103895 to int loses precision