Implicit conversion from float -4.857142857142861 to int loses precision