Implicit conversion from float -86.76923076923077 to int loses precision