Implicit conversion from float -68.70535714285714 to int loses precision