Implicit conversion from float -20.726643598615908 to int loses precision