Implicit conversion from float -17.94642857142857 to int loses precision