Implicit conversion from float -0.8018867924528195 to int loses precision