Implicit conversion from float -0.39473684210526017 to int loses precision