Implicit conversion from float -10.857142857142861 to int loses precision