> Yet, this tendency for myopia and prioritization of spectacle seems not align with many interests of mine. I find that a lot of the most significant levers on my life, both good and bad, seem to rely on compounding, on consistency and longer time periods.
This is quite insightful. It applies even more so to groups of people collectively (I'm sure I need not point out specific instances). All the more so when the data is noisy, and a bit of selectivity in setting the date range for analysis can result in the trend being minimized or reversed.
A moving average graph can help dispell this illusion, but the more aggressive the averaging the more it becomes a trailing indicator. One way to adjust for this is to use two moving averages (one longer one shorter) and plotting the difference between them. That will give you a fairly clear idea of whether the trend you're looking at is getting stronger, weaker, or reversing. It is still a trailing indicator but the trend-of-the-trend knowledge helps adjust for that.
You might want to look into my timeline thing [0]. At it's core, it's a database of Entries with different schemas.
- Input: There's an API you can add Entries with, and Sources that automatically pull them from somewhere.
- Output: There's an API you can query Entries from, and Destinations that automatically export them.
It's meant to be more like a diary and less like a dashboard, but once you have the data in a single database, it's easy to do other things with it.
A while ago, I made a map of my recent geolocation. It took maybe an hour, and allowed my dad to follow me during a trip. I wanted to make a maintenance schedule view for my vehicles, a budget view, and a few other things.
The cumulative distance of running since 1 January of each year is not particularly meaningful. Instead, the cumulative distance in the previous 365 days for each day would be a better metric. In such a diagram it would be easier to spot in which periods the performance is above or below avarge (or any other benchmark).
Love stuff like this, even tried logging a full year in Excel in 15min increments (a project seen here on HN), but nothing was ever as complete or automatable as I needed it.
I have Google Fit on my phone, I have a MiBand which tracks steps + heartbeat + sleep stats, is there a way to import these daily? And generate stats from them?
I've been working on a platform that allows you to log and track daily events https://www.simplejournal.online/
I've been more focused on collecting rather than processing the data and giving automated feedback, like what you're doing with your telegram bot. I really like that aspect. Very cool setup
I was very impressed by Felix Krause, who collected more than 380.000 data points over 3 years about his life in a single database and shared many of his learnings publicly: https://krausefx.com/blog/how-i-put-my-whole-life-into-a-sin...
Stressful year, Kevin. Hope it gets easier!
You can use Terra (tryterra.co) to access your garmin data !
undefined
I got myself a Garmin watch at the beginning of the year, to collect various metrics automatically. The watch uses the built in heart rate monitor/sensors to derive various data specifically:
- sleep hours (including the sleep phase type: deep sleep etc)
- stress amounts (via heart rate variability)
- energy levels ("Body battery" in Garmin speak)
I've been feeling quite drained the last couple of weeks so I wanted to see if the data I've collected over the last 3 months or so would match what I was subjectively feeling.
Interestingly Garmin does not provide any functionality to analyze long term trends, but there's an open source project to extract data from Garmin [0].
I used the tool to generate some graphs [1] that, do indeed, seem to indicate a rising level of stress over the last few months.
I'm going to try the moving average next to see if it's better than the naive approach I used, but ultimately my goal is the same as author's. I want a warning to sound off based on sleep/stress/energy levels trends. I have a tendency to overdo things sometimes. My theory is that a day off taken before some critical level is better than a week off after the burn out.
Here's the PR with the Jupyter notebook that generates the graph in the link based on Garmin Data [2].
[0] https://github.com/tcgoetz/GarminDB
[1] https://imgur.com/a/Q7MJqMB
[2] https://github.com/tcgoetz/GarminDB/pull/155