Implicit conversion from float -10.471698113207538 to int loses precision