Implicit conversion from float -50.78014184397165 to int loses precision