Implicit conversion from float -4.17613636363636 to int loses precision