A lot of modern so-called “advanced” basketball analytics, like RAPM, are attempts to ignore traditional statistical measures and use on/off or plus/minus stats to holistically capture performance. This method doesn’t work particularly well without secreting arbitrary, consensus-affirming weightings into the mix, but it’s growing in popularity anyway. Over at Boxscoregeeks, Patrick Minton has published a very simple set of thought experiments that show why on/off NBA analytics are flawed.