Implicit conversion from float -18.545454545454547 to int loses precision