Implicit conversion from float -39.10714285714286 to int loses precision