Regression to the Mean in Personal Data
Why unusually good or bad readings tend to be followed by more ordinary ones, and how this can make a harmless habit look like it worked or failed.
Content last updated . This page has not been medically reviewed.
The pattern
Regression to the mean (RTM) is a statistical phenomenon that can make natural variation in repeated data look like real change. It happens when unusually large or small measurements tend to be followed by measurements that are closer to the mean, meaning the average. [1] It is not a force pulling you toward average; it is what happens when a measurement includes some luck.This guide is information only. It is about reading your own data, not about any treatment.
A bedtime example
Imagine you have a run of poor nights and decide to try a calm wind-down routine. The next few nights are better. Was it the routine? Maybe. But very poor stretches are unusual, and the nights after one tend to look closer to your usual level whether or not you changed anything. You started when things were at their worst, so almost any change would seem to help.The same goes in the other direction. If you have an unusually good week, add a new habit and then slide back, you might wrongly decide the habit harmed you.
Walking and other measures
Suppose you notice your afternoon energy was very low on one day, then start taking a comfortable walk. If energy is better the next day, part of that may simply be the typical bounce back from an unusually bad day. Other personal numbers behave the same way. Daily resting heart rate, for example, usually remained relatively stable in one large study, but 20% of people had at least one week in which it fluctuated by 10 beats per minute or more. [2] A reading taken in one of those weeks would likely be followed by one nearer the person's usual.When RTM is strongest
Researchers note that the effect of RTM becomes more noticeable with increasing measurement error, and when follow-up measurements are examined only on a sub-sample selected using a baseline value. [1] In personal terms, that means:- Noisy measures (a single rating or a device estimate that jumps around) are more affected.
- Starting because of an extreme reading invites it. The very reason you chose to act makes your starting point unusual.
What this does not mean
It does not mean change is never real. It means one swing is weak evidence. And it is no reason to ignore a persistent or worsening problem: if your symptoms keep getting worse, or you have a concerning symptom, see a clinician. See talking with a health professional.Frequently asked questions
What is regression to the mean in plain words?
Unusually high or low measurements tend to be followed by ones closer to average, which can look like real change. [1]
Does it mean my improvement was not real?
Not necessarily. It means a single before-and-after comparison cannot separate a real effect from a natural bounce back. [1]
Can a tracker's score bounce back too?
Yes. Any repeated measurement with natural variation or measurement error can show it. [1]
How can I reduce its effect?
Use a baseline of several ordinary days, avoid starting because of one extreme reading, and look for repeated differences rather than a single one. See changing one low-risk variable at a time.
More details: evidence, limits and cautions
The details behind this page: cautions, limits of the evidence and where sources disagree. Open any note to read it.
Ways to reduce the risk of being fooled
- Build a baseline first with several ordinary days rather than reacting to one bad one. See establishing a baseline.
- Do not judge a change on one or two readings. Look at the overall pattern and its usual range; see baseline variation and normal fluctuation.
- Consider switching back and forth. Researchers describe N-of-1 trials as observing outcomes under periods of an intervention and a comparator. [3] Seeing a difference more than once is more convincing than one before-and-after.
- Be honest about what you cannot rule out. RTM is described as ubiquitous in repeated data and something that should always be considered as a possible cause of an observed change. [1]
References
- Barnett AG et al. Regression to the mean: what it is and how to deal with it. Int J Epidemiol 2005. PubMed 15333621. Read 2026-10-22.
- Inter- and intraindividual variability in daily resting heart rate and its associations with age, sex, sleep, BMI, and time of year. PLOS ONE 2020. PMC7001906. Read 2026-10-22.
- Individual-centric N-of-1 trials: a case study assessing the effect of alcohol abstinence on mood levels, 2026. PubMed 41501684. Read 2026-10-22.