This is even crazier than my last crazy idea, but what if you could define rate and max buffer using dynamic formulas?
rate = avg( day_vals )
max_buffer = avg( day_vals ) * 2
max_buffer = max( day_vals )
rate = avg( day_vals ) * 1.1
rate = avg( day_vals ) * .9
rate = max( avg( day_vals ) * 1.1, 10 )
rate = min( day_vals ) * 2
rate = avg( points ) * 3
Does any of that look useful, or am I suggesting madness? 
The idea of having your rate automatically set is something we’ve been calling autodial.
I don’t think autodial is madness, but I think supporting everything you’ve got there in an initial release may be 
Some of the ideas I’ve heard for basic use cases like “if my two week average rate goes higher than the current rate, set the rate to that, unless it’s past some defined maximum” or “just like that, but increase the rate a little bit more to push me (or less, depending on which way the road is going)”
I’ve got something like rate = avg( day_vals ) with http://maryrenaud.com/beeminder/mary-autodial.php (SUPER slow to load, cause when you load it, it dials all my graphs!) It’s an autodial thing of the type @adamwolf mentioned.
If you’d rather see an example screenshot than load the slow page, it looks like this:
It takes the average of the last 100 days and makes that my rate requirement.
(Self conscious note: For graphs that don’t have 100 days yet, or have been recently restarted, the requirements are going to be very low.)
This is super interesting. Thank you for sharing!
One thing that’s not clear to me: Does this approach mean that the rate would never get easier, and only harder? Or are there circumstances where the rate would become easier?
Also, do you have any goals you don’t auto dial?
TL;DR: Without interference or breaks, yes, it would keep getting harder.
The rate could get easier if I set it explicitly or took a break or something like that. The autodialer that I made only changes the final segment of the road, so I could insert a long break somewhere to slow things down if I needed to, and that break wouldn’t get destroyed (unless I had a max safety buffer set). Without that, though, it should mostly only get more challenging until it hits the goal rate. And that can be sped up somewhat by having the safety buffer automatically trimmed. The key, though, is that it never gets harder than I’ve already managed to pull off for 100 days, so it’s a nice slow incline towards my ultimate goal rate, and the proof that I should be able to maintain it (so long as nothing relevant changes) is that I’ve already been doing it. (…Unless the safety buffer is set low enough that it’s triggering a requirement to do the thing faster than the set rate itself would, which will actually be the case with some of these.)
TL;DR: Yup, almost half of them aren’t, but a lot of them are tests or weird kinds of goals.
Yup. In fact, most of these are recently restarted goals that had been archived for a while while I took a step back to think about how I wanted to organize things. I have about 90 Beeminder goals (some of which are just test goals, others of which include meta goals, meaning there are multiple Beeminder goals to 1 real-life goal; my intermittent fasting goal, for example, uses 4 Beeminder goals.) Anyway, of those 90, about 50 have now been set up to autodial again and the other 40ish aren’t.
