don't extrapolate the turquoise swath into the future
A few days might be OK, but definitely not two weeks. It can be silly, especially when the polynomial is high-order: see e.g. https://www.beeminder.com/army1987/goals/weight -- I don't think I'm gonna weigh 92 kilos by the 28th, not even taking the Christmas holidays into account...
Yes, that's OK now. (If you do want to extrapolate that far into the future, you might want to use a better algorithm such as a lower-order polynomial (or some other function entirely) and/or weight recent data points more than ancient ones.)
Admindreeves (Cofounder, Beeminder) commented
Thanks for pointing this out! Sometimes it's nice when it's actually extrapolating reasonably, but you're totally right.
Do you think it's less egregious when it just extrapolates up to the akrasia horizon? (See your graph now.)