Implicit conversion from float -51.711590296495956 to int loses precision