Implicit conversion from float -72.87878787878788 to int loses precision