Implicit conversion from float -203.38235294117646 to int loses precision