Implicit conversion from float -169.625 to int loses precision