Implicit conversion from float -52.85714285714286 to int loses precision