Implicit conversion from float -19.83471074380165 to int loses precision