Implicit conversion from float -103.35664335664336 to int loses precision