My goal rate here is 70/week (recently increased from 50, hence the bent road). I can eyeball the rate at which I am doing pullups and see i’m doing OK, but I’d like to see by how much I am beating the rate; in particular, I’d like to see a data point somewhere on the page “Current rate: ~85/week” (or whatever it really is.)
The way I’d do this, to be more concrete (though I’d be open to other implementations) would be a moving average–either take a window of a week or so and compute a rate, or (possibly better, but I’d have to tinker to find out) do an exponentially weighted moving average with slow decay on historical interarrival times. For lots of extra points, compute some form of approximate error bars, and graph the cone of my current progress on my graph so I can see how it compares to my YBR.
Getting this to work for odometers/weight loss would be trickier (you’d have to take a derivative or similar at some point) but starting out showing this for do more/do lesses would be more than enough.
Ooh, this would be great. It would also be useful when you start a goal but don’t know what the rate should be. You can set a very low rate and collect data for a while, and then have Beeminder automatically give you an idea of the rate you have actually been achieving, which you can use to decide how to set the rate.
If you tick the moving average box on https://www.beeminder.com/ahh/goals/pullups/graph_settings (that’s advanced settings for the rest of you!), you’ll get a visible exponentially weighted moving average skinny purple line on your graph. That might help with the eyeballing.
I’m also hopeful that the (forthcoming, eventually) Road Editor will serve this purpose somewhat. You’ll be able to keep retroactively adjusting the yellow brick road to match your data – trial-and-error until you have something that feels right. In a lot of ways that will be better than sophisticated smoothing/extrapolation, though I’m game to do more of the latter too!
Philip–the moving average on the graph is very nearly what I want (and is definitely useful). But I need it to either display the numerical rate estimate or (better yet) extrapolate out into the future of the graph.