Implicit conversion from float -15.6949806949807 to int loses precision