Implicit conversion from float -102.85714285714283 to int loses precision