Implicit conversion from float -1.4285714285714164 to int loses precision