Implicit conversion from float -51.81141439205956 to int loses precision