NOTES
Here I fit the data with a 15th-order polynomial. Those of you who fit data to models regularly already know what's wrong with this, but for now just notice that the amount of noise is somewhat smaller - our theory explains more of the data, but uses more parameters to describe the fit.