I like data. Probably a little too much.

If there is a number that promises to tell me how recovered I am, there is a very good chance I will look at it. HRV, resting heart rate, sleep duration, readiness, recovery score. Give me a dashboard and I will happily stare at it before I have even worked out how I actually feel.

That has more or less been my morning routine for a while. I wake up, open the health app and look at the report. Green feels reassuring. Yellow is enough to make me pause. And before I have really noticed whether my legs feel heavy, whether I slept well or whether the slight scratch in my throat actually means anything, I am already interpreting the number.

Why is my HRV lower? Did I sleep badly? Was yesterday harder than I thought? Am I getting sick? Was it just a bad measurement? Did I sleep on my arm weirdly? Should I change today’s training?

The funny thing is that I often start asking those questions before I have asked the much simpler one: how do I actually feel?

A number that was supposed to reduce uncertainty has somehow created six new things to think about before breakfast. And I know I do this. I still look.

I used Whoop for quite a while, and I completely understand why people like it. It takes something messy and turns it into something very clean. How recovered am I? 72%. Green. Great.

There is something deeply satisfying about that, especially if you are the kind of person who likes measuring things in the first place. Which, unfortunately for this experiment, I very much am.

I like dashboards. I like trends. I like turning vague feelings into something I can look at. At some point, though, my recovery dashboard started to feel a bit like a tiny Tamagotchi I had to keep green.

A bad score did not just tell me something about last night. It changed how I thought about the day. Maybe today’s workout was suddenly too much. Maybe the slight scratch in my throat was actually something. Maybe I had not recovered from the last session after all. Maybe I should check the number again, just in case it had somehow improved in the last three minutes.

The problem was not necessarily that the score existed. The problem was how badly I wanted it to be right.

I do not think HRV is useless, and I do not think sleep data is useless either. I am also definitely not about to become the guy who tells everyone to throw away their watch and simply listen to their body. That would be a fairly dramatic character development.

I still want the data. I just do not want every measurement to become a verdict.

Maybe the problem is not measuring recovery. Maybe the problem is that I try to turn every single measurement into a decision.

One unusually low HRV reading does not automatically mean something is wrong. One great night of sleep does not make me indestructible. And one green recovery score definitely does not know everything that happened to me yesterday, especially when some of that “everything” is carrying things around a building site, sleeping badly, working late or just having a stressful day that no wearable really understands.

That is also why I do not want to build a better recovery score. The score is not the part I am missing.

What I actually care about is whether something is changing. Is HRV drifting down over several days? Is resting heart rate consistently higher than normal? Am I sleeping less than usual? Do several of those things happen at the same time?

That feels much more useful than asking whether today is green or yellow.

So the protocol I am trying now is simple: collect daily, interpret trends.

I still measure HRV, resting heart rate and sleep duration, but instead of looking at each morning in isolation, I want to compare a short rolling trend with a longer personal baseline. Right now that means roughly a seven day rolling trend compared with the previous several weeks of my own data.

Seven days is not some magical physiological window. It is simply far enough away from one weird night to stop me from immediately constructing a story around it. At least that is the theory.

The obvious problem is that knowing I should ignore daily values does not actually stop me from looking at them. My reflex is still there. Wake up, open app, check tiny digital version of myself.

So I started thinking about the interface rather than the data.

What if I simply did not see the daily value? What if the system kept collecting everything in the background, but stayed quiet unless the trend actually moved outside my normal range?

That is the version I want. I do not need another dashboard that gives me one more reason to check in every morning. I already have enough numbers. What I want is something that quietly watches the trend and only interrupts me when something actually changes.

If HRV, resting heart rate and sleep all look normal, nothing happens. No notification, no “everything looks great today”, no new little circle to keep green.

If the seven day HRV trend moves meaningfully below my recent baseline, I get a heads-up. If several recovery signals move outside their usual ranges together, I get a heads-up. That is it.

No automatic “take a rest day”. No “you are 43% recovered”. Just: something changed, maybe pay attention.

The default output is silence.

The slightly ridiculous part is that I already had most of the plumbing. My health data ends up in Intervals.icu anyway, so instead of building a full app, I set up a scheduled check that looks at the recent wellness data, calculates the rolling trends and compares them with my personal baseline.

If nothing looks unusual, it does absolutely nothing. If something does, I get a recovery heads-up.

That is currently the entire product, which is probably good.

I did not need another recovery score. I needed a system that would leave me alone unless something actually changed.

Good recovery monitoring might be something I mostly do not see.

And yes, I am apparently now using automation to stop myself obsessively checking automation. That feels about right.

I do not know yet whether this will actually work. The thresholds might be too sensitive. They might be too conservative. The whole thing might just become a slightly more sophisticated Tamagotchi.

That is why I am treating it as an experiment, not a finished recovery model.

For now, I want to know one thing: can I keep collecting useful recovery data without letting every morning’s number decide how I feel about training?

If the answer is yes, that would already be enough.