Implicit conversion from float -26.218487394957975 to int loses precision