Implicit conversion from float -35.20408163265307 to int loses precision