Implicit conversion from float -88.7037037037037 to int loses precision