Implicit conversion from float -68.24175824175825 to int loses precision