Implicit conversion from float -0.714285714285694 to int loses precision