Implicit conversion from float -10.57142857142857 to int loses precision