Implicit conversion from float -26.57142857142857 to int loses precision