A pound is not 3,500 calories.

Where the number came from

In 1958 Max Wishnofsky published a short paper in the American Journal of Clinical Nutrition working out the energy held in a pound of body tissue. He took an early analysis putting human adipose tissue at about 87% fat, valued that fat at 9.5 calories per gram, and arrived at roughly 3,750 calories in a pound. He then checked it against a 59-day study of people on a very low calorie diet, got 3,500, and called the agreement striking.

For what he was doing, it was reasonable work. The trouble is what happened to it afterwards. A figure describing the energy content of tissue already lost got turned into a formula for predicting tissue you have not lost yet, and those are different claims.

What the rule quietly assumes

Reading it forward as a prediction commits you to four things at once, and each of them is false to a degree. That your intake stays exactly where you set it. That your expenditure does not change as you get smaller. That what you lose is essentially all adipose tissue. And that the energy in a pound of lost weight is the same in week one as in month six.

That last one has been measured. Thomas and colleagues, writing in the Journal of the Academy of Nutrition and Dietetics (2014), report the energy content of weight change rising from about 4,858 calories per kilogram at four weeks to about 6,569 at twenty-four, still well under the 7,716 the rule assumes, and moving the whole time. Early weight is cheap because a good deal of it is water and glycogen. Later weight costs more.

How far off it gets

Far enough to matter within the first year. Hall and colleagues, in The Lancet (2011), simulated a 100 kg sedentary man cutting about 480 calories a day. The 3,500-calorie rule predicts a straight line down to 22 kg lost in twelve months. Their model, built on doubly labeled water data, predicts about half that.

The same paper is blunt about why the rule cannot describe what people actually experience: it draws a line, and a line has no plateau. Their own simulation settles at a new weight after several years, reaching half the total change in about a year. As they put it, practitioners of the rule generally knew loss would slow, but “had no way to estimate the weight-loss time course.”

The gap is not academic. Thomas and colleagues ran the classic one-can-of-soda example (140 calories a day removed from a 76 kg woman) and the rule promises 15 pounds in a year against the 6 to 8 the dynamic models give. Against measured data from a controlled feeding study, the rule was out by 4.8 kg at twelve weeks and 11 kg at twenty-four.

Why the deficit per pound is not one number

Hall showed in the International Journal of Obesity (2008) that the energy deficit required per unit of weight lost depends on how much fat you carried to begin with. Above roughly 30 kg of body fat the old rule is about right; below that it overstates the deficit needed, because more of what comes off is lean tissue and lean tissue is far cheaper.

The numbers underneath that are stark. Body fat carries about 9,440 calories per kilogram; lean body mass carries about 1,820. So the composition of the loss changes the arithmetic by a factor of five, and composition is not fixed. Cava, Yeat and Mittendorfer, reviewing the evidence in Advances in Nutrition (2017), put fat-free mass at 20 to 30% of total weight lost in people who are overweight or obese and often more than 35% in people of normal weight.

Note also that body fat and adipose tissue are not the same thing. Adipose tissue is roughly 87% fat with fluid and protein making up the rest, which is how Wishnofsky got 3,750 rather than 9,440 in the first place. Much of the confusion around this number comes from those two being used interchangeably.

What sits in its place

Hall’s group offers a replacement rule of thumb in the same 2011 paper, and the useful part is that it is explicitly slow: every permanent change of 10 calories a day leads to an eventual change of about one pound, with roughly half of it arriving in the first year and 95% within about three.

Read that carefully, because it is easy to flatten back into the thing it replaced. It is not a pound a year. It is a pound eventually, most of it late. The same model runs inside the NIH’s Body Weight Planner, which is free and asks for a date rather than promising one.

What it means for a tracker

Three things follow, and none of them is that the arithmetic is useless.

A projected date is a projection. Any app that multiplies your daily deficit by the days remaining is running Wishnofsky’s rule, and it will be optimistic in a way that compounds. A forecast that slows down is not being pessimistic; it is being arithmetically honest.

Slowing down is the prediction, not the failure. The dynamic models say loss decelerates and eventually stops at a new weight for a fixed intake. When your own rate falls off after a few months, the model expected that.

Your own trend beats anybody’s equation. Every figure above is a population average with a real spread around it. Several weeks of consistent logging tells you what your body did, which no equation can.

How Loopa handles this

Loopa’s forecast runs on your own rate of change rather than on a fixed calories-per-pound constant, so it decelerates as you lean out, shows a plateau when your data says plateau, and declines to promise a goal date it cannot reach safely. What If lets you change an input and watch the projected date move before you commit to it. Logging food and weight is free; the forecast and What If are Premium.