Implicit conversion from float -22.631578947368425 to int loses precision