Implicit conversion from float -51.03840682788052 to int loses precision