Implicit conversion from float -121.84834123222748 to int loses precision