Implicit conversion from float -44.423076923076906 to int loses precision