Implicit conversion from float -0.33018867924528195 to int loses precision