Implicit conversion from float -103.96551724137933 to int loses precision