and fitting whatever low-order polynomial has the best R^2 becomes more than reasonable...
Not really. But let's assume that's true for the moment. That means that Beiting did it all wrong, because the quartic fits have better R^2 values. Glad you agree with me that B needs to up his game a bit, Z...;)
{edit: try starting here http://blog.minitab.com/blog/a…the-best-regression-model}