Implicit conversion from float -0.7196969696969688 to int loses precision