Implicit conversion from float -16.861313868613138 to int loses precision