Implicit conversion from float -18.717948717948715 to int loses precision