Implicit conversion from float -23.273195876288668 to int loses precision