Implicit conversion from float -51.40127388535032 to int loses precision