Exercise is a poor way to lose weight — and the best way to keep it off

A physiologist's essay has been going round again these past two weeks with a claim that sounds like a demotion and is actually a promotion: exercise is not much good for losing weight, but it is very good for not putting it back on. The comment sections did what comment sections do. Half the readers heard the gym is pointless. The other half heard see, I knew running was a scam.

Both misread it. The finding is not that exercise fails. It is that we have spent forty years grading it on the wrong exam.

The number that made exercise look useless

Start with the unflattering evidence, because it is real and it deserves its moment.

In 2009 a team led by Tianying Wu pooled the long-term trials that had asked one clean question: if you add exercise to a diet, how much more weight do people lose? The answer, at long-term follow-up, was about 1.14 kg (95% CI 0.21–2.07) — roughly two and a half pounds, on top of whatever the diet did on its own.

A year of training. Two and a half pounds. That is the entire measured return, and it is why the "exercise doesn't cause weight loss" line keeps resurfacing every eighteen months dressed as a revelation.

The arithmetic behind the disappointment is dull and stubborn. A hard hour buys a few hundred calories. A moderately generous afternoon spends them back without you noticing, because appetite is not a spreadsheet and the body quietly economises elsewhere — fidgeting less, moving less between sessions. If you are looking for weight to fall off the scale, exercise is a slow, expensive lever.

The same habit, scored in a different phase

Now change the question. Not does exercise help you lose it, but does exercise help you hold it.

In 2026 a group led by J. Wang ran a systematic review and meta-analysis of trials that started after the weight came off — eleven randomised trials, 568 people, all of them in the maintenance phase. The exercise groups regained 2.81 kg less than the controls (95% CI −5.12 to −0.51). The heterogeneity statistic was I² = 0%, which is unusual and worth pausing on: the trials mostly agreed with each other.

So the same behaviour, in the same units, on the same scale: about a kilogram when you ask it to remove weight, and nearly three when you ask it to defend weight. It is not a weak intervention. It is a well-aimed one that we keep pointing at the wrong target.

Why a body defends a weight it has already lost

The best-known evidence for how hard that defence gets is also the most quoted and the most contested. A team at the US National Institutes of Health — Erin Fothergill, Kevin Hall and colleagues — followed contestants from a televised weight-loss competition for six years. They had lost an average of 58 kg. Six years later they had regained 41 kg of it — and their resting metabolic rate was still running 704 kcal/day below baseline, far lower than their regained body size could explain.

That study did enormous damage to the idea that maintenance is a matter of willpower. It also did some damage to the truth, because it has been repeated ever since as though it were the settled physiology of everybody.

It isn't. A 2020 analysis in the American Journal of Clinical Nutrition by Catia Martins and colleagues argued the opposite reading: that metabolic adaptation is largely a feature of active energy deficit, shrinks once weight is stable, and is not the major barrier to keeping weight off. And the famous cohort was fourteen people under conditions almost nobody replicates — the case for caution is decent.

I find both readings survivable, and the honest position is the boring one: the body does push back, the size of the push is disputed, and the dispute doesn't change what you do on a Tuesday. It is worth reading our piece on what actually happens to metabolism with age for how easily this particular number gets mythologised.

What the people who actually kept it off do

There is one more body of evidence, and it is the least rigorous and the most human.

The National Weight Control Registry has spent decades tracking people who lost a lot of weight and kept it off — an average of about 33 kg held for more than five years. Rena Wing and Suzanne Phelan's summary of what those people report is almost aggressively unglamorous: they eat a consistent lower-calorie diet, they eat breakfast, they weigh themselves, and they do roughly an hour a day of physical activity.

An hour a day is a lot. It is also, in this evidence, the most consistent single behaviour among the people who succeeded — and only about one in five people who lose 10% of their body weight are still holding it a year later.

The obvious caveat is the fatal one: this is a self-selected registry of winners. Nobody enrols the people it didn't work for. It cannot tell you that the hour causes the maintenance rather than accompanying it — the kind of person who can protect an hour a day may be protecting several other things too. Treat it as a description of a successful life, not a prescription.

Exercise doesn't burn the food. It changes the negotiation.

If exercise isn't working by simple subtraction, what is it doing?

Two mechanisms have decent support. The first is that training protects muscle while weight comes off, so more of what you lose is fat and more of what remains is metabolically busy tissue. The second is stranger and more interesting: appetite regulation appears to work better in people who move a lot. A 2025 review in the European Journal of Clinical Nutrition sets out the idea of a "regulated zone": at higher habitual activity, hunger tracks energy expenditure more tightly, and intake and output stay coupled. Sit still all day and that coupling degrades; appetite stops corresponding to anything in particular.

Which inverts the folk model completely. Most people treat exercise as the thing that pays for the eating. The evidence suggests it is closer to the thing that keeps the eating legible — that makes hunger informative again rather than ambient.

None of this is a licence to promise anyone an outcome. These are averages from trials with dropouts and mechanisms mapped in modest samples. But the direction is consistent enough to act on.

The scale is the wrong instrument

Here is the part that annoys me, as someone who builds measurement tools for a living.

The bathroom scale won this argument by default, because it was the only number in the house. It is a genuinely bad instrument for what exercise does: it cannot tell muscle from fat from Saturday's salt, it moves two kilos on water alone, and it reports nothing at all about the two things training actually changed this month — how much work you can do, and how well you recover from it.

So if you train and the scale doesn't move, the scale is not delivering bad news. It is declining to comment.

Things that do comment, and that you can watch over weeks rather than days:

• Load you can handle. Reps, weight, hill, pace held. The cleanest evidence that the last two months existed. See what you lose when you stop training for how fast that unwinds.
• Waist, not weight. A tape measure at the navel tracks the fat distribution that body weight hides — see waist size versus BMI.
• How quickly your heart rate falls after a hard effort — the trend, not any single reading (heart-rate recovery).
• Strength you can test at home. Grip is the cheap proxy that keeps showing up in ageing research (grip strength and longevity).

This is what the Agen Band and app are built around: not a verdict each morning, but a set of trends slow enough to be true. Numbers are there to correct the fantasy, not to replace the experience of the week. If you want the whole picture assembled, start with building a longevity protocol or how much exercise you actually need per week.

The bottom line

Exercise is a poor subtraction tool and an excellent holding tool, and the research finally has the numbers to say so: about a kilogram when it is asked to strip weight off, close to three when it is asked to keep weight from coming back.

The practical consequence is small and unromantic. Do not start training in order to watch the scale fall, because it mostly won't, and the disappointment is what ends the habit around week seven. Start it because it is the thing that makes any change you do manage survivable — and then judge it on load, waist, recovery and strength, which are the four things it is actually moving.

The exercise was never underperforming. The scoreboard was.