You might remember that I implemented a thing for beeminder that I’m calling backpressure. If a goal has less than a week to derailing it generates a datapoint each day. There’s then a “do less of that” goal.
It required a bit of tinkering at first: My initial rate was really low compared to what I actually generated on this front, so I had a few days of scrambling panic to get it under control. Additionally I realised it only worked well for “do more” goals where I can actively work towards making things better as opposed to just not doing anything to make that goal worse for a few days.
Once I got things under control I also decided that my intended rate was too generous, and dialled it down to one which didn’t allow me a full week of leniency: If I have a permanent rate of about one a day being inside the backpressure window than I will derail.
The result is pretty great. I no longer have eep days, I have “oh dear” several day periods (needs a catchier name). When something enters the backpressure window I tend to immediately bat it out again by doing some extra work on it, but it’s no big deal if I need to leave it a day or two. It’s an altogether much calmer experience.
But there is one slight case where it fails to be calmer: Because most of my goals tend to hover a bit outside the one week window now, I still suffer from the correlated eeping problem where multiple things enter the backpressure window at once and I have to deal with both of them because several things in backpressure will cause me to derail much faster. It’s still not nearly as bad as having multiple simultaneous eep days, but it could be better.
I have a couple ideas for how to deal with this, but the easiest one involves me going “Hmm… I’ve got this hammer in my hand. I wonder if I can solve this problem by hitting things?”
So the solution to backpressure’s problems? MORE BACKPRESSURE.
I’ve created a second goal, backpressureharder, which tracks how many goals I have with less than two weeks of grace period left. The idea is that this will be different because it will have a much laxer rate. My current plan is to keep it to just under 2/day, so that I can safely let a single goal slide into the full blown backpressure window, but if two are in there at once I need to deal with it or I’ll derail (or at least burn through my bufffer) – it shouldn’t derail fast. I don’t want to be frantic about it, but at some point in the week I should do something about it.
My initial goal setting on it is not anything close to that because I want to avoid the frantic scrambling around of last time, so I’ve actually got it set to 10 a day. My plan is to clear things out of it gradually, adjust the road dial downwards and retroratchet aggressively to keep myself honest.
P.S. No I’m not just writing this blog post to get my blogging goal closer to being outside the two week window. Why do you ask?
I predict that as t -> \infty, the number of recursive backpressure goals in your system will grow unboundedly.
I wonder the same but I don’t think so. It isn’t very useful past this point and if I need something more it’ll be time to try something else