Implicit conversion from float -91.0344827586207 to int loses precision