Implicit conversion from float -26.5263157894737 to int loses precision