Implicit conversion from float -304.7368421052632 to int loses precision