Implicit conversion from float -98.65546218487395 to int loses precision