While standard regression models are great at many things, they struggle when one x leads to several valid y values—they often just "split the difference" and predict an average that isn't actually ...