Implicit conversion from float -21.195652173913032 to int loses precision