It seems to me that the polynomial fit trend doesn't make sense with a weight loss goal. Beeminder links to the Hacker's Diet, and his (now accomplished) weight loss pattern is a hockey stick; straight line decline, and then horizontal maintenance. The best fit to a successful weight loss program might possibly be exponential (?), but I'm not sure curve fitting adds anything, and it might make the graph too noisy.
I'm satisfied with my progress to date, but my current graph shows an uptick in the polynomial fit. My sense is that short term curves are meaningless, and long term weight loss graphs are so variable that there's no added value in having it on the graph. As long as I can stay below both the moving average and the yellow brick road, I'm happy.
(Maybe an option to suppress the polynomial fit trend?)