I think if '52-48' is redefined as a 'big margin' then it becomes very hard to define 'a narrow margin' without running into the impact of statistical noise.