Energy Balance for Active People

Calories in versus calories out is true. Body mass changes when energy intake and energy expenditure stop matching, and that has been confirmed in metabolic chambers, doubly labeled water studies, and decades of dynamic modeling. The reason most active people still struggle to make the math work is that both sides of the equation move under their own control loops. Intake is mismeasured. Expenditure is variable. Training changes appetite. Dieting changes spontaneous movement. The first law of thermodynamics still holds. The day-to-day accounting is what's noisy.

This guide is for the person who already knows that calories matter and is now trying to understand why a 500 kcal deficit on paper produced no scale movement, why a hard training week drove hunger but flat weight, or why the same intake at the same body weight feels different a year apart. The goal is to lay out the full model, name each component, and give you a calibration loop that respects how noisy each piece really is.

01What energy balance actually says

Energy balance is the relationship between energy intake and total energy expenditure across a meaningful window of time. When intake exceeds expenditure, the surplus is stored, mostly as fat with some lean tissue if a training stimulus is present. When expenditure exceeds intake, the body draws on stored energy, again mostly fat but with some lean tissue at risk during aggressive deficits.

Hall and colleagues at the NIH built and validated a dynamic mathematical model of this relationship across multiple controlled-feeding studies, showing that long-term weight change in adults is well predicted by the integrated difference between intake and expenditure once the adaptive responses on both sides are accounted for.¹ The same group has shown that the older "3,500 kcal equals one pound of fat" rule overestimates real-world fat loss because it ignores the predictable downward drift in expenditure that accompanies a sustained deficit.²

The first law is preserved. The complication is that intake and expenditure are coupled through behavior. A larger deficit usually drives lower spontaneous movement, higher hunger, and worse sleep, all of which feed back into both sides of the equation. Treating energy balance as a static input-output box misses the loop. Treating it as a dynamic control problem with measurement error on every channel matches reality.

02The expenditure side has one stable piece and three behavioral ones

Total daily energy expenditure is the sum of all energy your body uses across 24 hours. It breaks into four components with very different magnitudes and very different stability across days.

For an active 80 kg adult, a representative breakdown might be RMR 1,750, TEF 220, EAT 350, NEAT 450. Total around 2,770 kcal per day. The same person on a sedentary travel day with a missed session might land at 1,750 RMR, 200 TEF, 0 EAT, 200 NEAT. Total around 2,150. The 600 kcal swing between those days is real, and it is almost entirely behavioral.

RMR is the base, and it is more stable than most people assume

Resting metabolic rate is the energy you spend at rest while awake in a relaxed state. It is the largest single component for almost everyone, and it is dominated by the cost of running high-metabolic-rate organs. Heymsfield's organ-tissue work showed that the brain uses roughly 240 kcal per kg per day, the liver about 200, and skeletal muscle only about 13.³ Most of the variance between healthy adults of similar weight collapses once you control for fat-free mass and organ mass. Resistance training supports RMR less by raising the metabolic cost per kg of muscle and more by preserving total fat-free mass during a cut.

For estimation, the Mifflin-St Jeor equation is the most defensible single formula in non-obese and obese adults, predicting within 10 percent of measured RMR in 70 to 82 percent of subjects.⁴ That accuracy ceiling matters. For roughly one in five users, the equation is off by more than 10 percent, which is enough to move daily intake recommendations by 200 to 400 kcal. Treat the equation as a starting estimate. The 14 to 28 day weight trend is the calibration. The distinction between BMR and RMR is small in practice. Most equations marketed as BMR formulas are actually calibrated against RMR data.

TEF is real, useful for design, and oversold

The thermic effect of food is the energy your body spends digesting and processing meals. On a mixed diet, TEF is roughly 10 percent of intake. It is not uniform across macronutrients. Protein costs about 20 to 30 percent of its calories during processing, carbohydrate about 5 to 10 percent, and fat 0 to 3 percent.⁵ On a 2,500 kcal day, a higher-protein pattern usually produces 30 to 70 kcal more TEF than a lower-protein pattern at the same total calories. That is genuine, and it accumulates across a year. It is also smaller than the error created by a single weekend of unlogged restaurant meals.

The practical use of TEF is to design the diet, mostly by setting protein high enough to support both lean mass retention and satiety. The misuse of TEF is to treat it as a meaningful explanation for stalled progress. If your scale is flat for three weeks, the answer is almost never "my TEF is lower than it should be." It is usually intake drift, NEAT drift, or both.

EAT is the part most people overestimate

Exercise activity thermogenesis is the energy cost of planned training. For most active people, this lands somewhere between 200 and 600 kcal per session, depending on duration, intensity, body mass, and modality. Two patterns are worth naming carefully.

First, watch-reported active calories are not the same as EAT. The active calorie ring on a wearable folds in a chunk of NEAT alongside formal training, because the device cannot reliably tell standing-and-pacing apart from light cardio. That makes active calories a useful daily activity proxy and a poor input for "how much did this workout cost." See Wearable Calorie Accuracy for the broader research summary.

Second, the relationship between training time and added expenditure is not perfectly additive. Pontzer's constrained total energy expenditure model and later doubly labeled water work show that high activity volume can produce less added TDEE than the math of session calories alone would predict.⁶ The body may partially compensate by lowering NEAT, reducing fidget volume, and improving movement efficiency.

That model is debated in the way useful models usually are: the direction of compensation is well supported, the exact size varies by population, baseline activity, energy availability, and training history. A sedentary person who starts walking daily may see a large TDEE increase. A lean endurance athlete adding more volume during a deficit may see more of the new work absorbed by lower spontaneous movement and recovery behavior. The practical rule is not "exercise stops counting." Exercise counts, but the body adjusts other expenditure to partially offset it.

NEAT is the variance engine

Non-exercise activity thermogenesis covers everything that is not RMR, TEF, or planned training. Walking, standing, climbing stairs, carrying things, fidgeting, pacing during phone calls. NEAT is the most underappreciated component of TDEE because it is invisible inside any calculator that uses an activity multiplier, and it is the single largest source of person-to-person variance in expenditure.

Levine, Eberhardt, and Jensen's overfeeding study at Mayo Clinic gave 16 non-obese adults an extra 1,000 kcal per day for eight weeks. NEAT changes ranged from a decrease of 98 kcal per day to an increase of 692 kcal per day across subjects, and those NEAT changes accounted for the majority of inter-individual differences in fat gain.⁷ Two people eating the same surplus can gain very different amounts of fat depending partly on whether their bodies upregulate or suppress spontaneous movement.

The same dynamic runs in reverse during a deficit. People who lose weight tend to walk less, fidget less, and stand less. Trexler, Smith-Ryan, and Norton's review describes this as one mechanism inside metabolic adaptation that makes long cuts progressively harder.⁸ On a practical level, the difference between a 4,000-step day and a 12,000-step day is often 300 to 500 kcal of real expenditure. Step count drift is usually the first thing to look at when a previously reliable target stops producing scale movement.

03The intake side is noisier than people think

Energy balance fails in practice on the intake side at least as often as on the expenditure side. Three sources of error stack on top of each other, and most users underestimate how large the combined effect can be.

Food labels in the US have compliance tolerances that can allow meaningful variance from stated values. For calories and several nutrients, FDA compliance generally treats a product as misbranded when measured content exceeds the declared value by more than 20 percent, rather than promising a clean symmetric error band. Multiply that across an entire day's mix of packaged foods, restaurant menus, and home-cooked meals with imperfect ingredient databases, and the realistic error band on a "2,200 kcal day" can be wider than most users expect.

Self-reported intake compounds the problem. Lichtman and colleagues' classic study of self-reported diet-resistant subjects found under-reporting of intake averaged 47 percent and over-reporting of physical activity averaged 51 percent in people who believed they could not lose weight.⁹ Schoeller's doubly labeled water work confirmed the same pattern across many populations. Most under-reporting isn't dishonest. People forget the cooking oil, bites off a partner's plate, the second pour of cereal, and the cocktail at the work dinner.

Logging method changes the magnitude. A consistent pattern across studies and field practice:

The fix is not to chase clinical accuracy at home. The fix is to know which band you are in, accept that the absolute calorie number on your log is probably 5 to 20 percent low, and stop using that exact number to override a clear weight trend. If you logged 2,400 kcal per day for two weeks at stable weight, your true maintenance is approximately 2,400, not whatever the equation said and not whatever the watch said. Logging error is forgiven by the trend math when logging is consistent.

04The energy balance equation in practice

The equation as physics is simple.

Δ body energy stores = energy intake − total daily energy expenditure

The equation in practice has feedback on every term.

Δ stores = (intake × 1/log_accuracy) − (RMR_t + TEF_t + EAT_t + NEAT_t)

where RMR_t falls slowly with weight loss and may be suppressed further by adaptation, TEF_t follows intake and macro composition, EAT_t is partly compensated by changes in NEAT, and NEAT_t drops as the deficit deepens. None of those terms are knowable to better than 10 to 20 percent at the daily level. All of them are knowable to better than 5 percent at the multi-week level if you collect honest data and let the trend speak.

That is the whole reason calibrating off the trend works. The errors on every channel are roughly stable in direction and magnitude over time, so they cancel inside a 14 to 28 day window. The trend reveals net energy balance even when no individual day's accounting is exact.

05Maintenance, deficit, and surplus as relative numbers

Maintenance calories are the intake that holds body weight stable across a multi-week trend. A calorie deficit is intake below current maintenance. A calorie surplus is intake above. The point worth holding onto is that every one of these numbers is relative to a maintenance value that itself moves with body mass, training load, sleep, and dieting state.

The maintenance you had at 200 lb is not the maintenance you have at 180 lb. The maintenance you had during a heavy training block is not the maintenance you have during a deload week. The maintenance you had before a 16-week cut is rarely the maintenance you have at the same body weight after the cut, because adaptive thermogenesis can sit 100 to 200 kcal below body-composition predictions for weeks to months after restriction ends.¹⁰ In severe cases, like Fothergill's six-year follow-up of contestants from The Biggest Loser, the gap can persist much longer.¹¹

Reasonable starting bands for active adults using a Mifflin-St Jeor RMR with an honest activity multiplier:

Garthe and colleagues showed the cost of moving too fast on a cut. Elite athletes in the slower-loss group lost about 0.7 percent per week and increased lean body mass by 2.1 percent, while the faster-loss group lost about 1.0 percent per week and left lean body mass roughly unchanged at -0.2 percent.¹² Aggressive deficits work for a few weeks and then often collapse through some combination of poor sleep, declining lifts, and rebound eating. Slower cuts compound.

06How training changes the math

Adding training to an inactive baseline does raise TDEE, just less than a naive sum suggests. Three patterns deserve attention because they explain a lot of stalled cuts and unexpected scale moves.

First is the constrained expenditure pattern. Adding structured training raises TDEE, but usually by less than the full session-calorie sum once NEAT, movement efficiency, and recovery behavior adapt. The pattern is more pronounced in well-trained populations and during deficits. The practical reading is that calorie targets built from "BMR times 1.55 plus per-session calories" tend to overshoot the actual TDEE on heavy training weeks, especially when training is added on top of an existing high-volume block.

Second is the appetite response. Hard training, especially long cardio, raises post-session hunger out of proportion to the energy actually burned. King and colleagues' work on exercise-induced appetite changes shows wide individual variance, with some subjects compensating fully through eating and some not compensating at all.¹³ If you are someone whose appetite tracks your training load, your weekly intake will tend to creep up during high-volume blocks even without conscious change. Build that into the plan instead of fighting it weekly.

Third is the recovery cost. Training expenditure does not stop when the workout ends. Heavy resistance training carries a small excess post-exercise oxygen consumption tail, and the muscle protein synthesis response after a hard session adds a small amount of low-intensity background expenditure. Both are too small to budget for individually. Both contribute to why active people who feel hungry after lifting often are.

07What goes wrong without a calibration loop

Most stalled cuts and unexpected weight gains in active populations come from one of a small set of failure modes, all of which trace back to treating a static number as if it were a measurement.

Adaptive Calorie Goals handle most of these failure modes by holding intake stable until two or more independent signals agree on a change. Adaptive targets move slowly on purpose, because reactive systems chase noise.

08Calibration is the only loop that actually works

A defensible calibration loop has four moving pieces and one decision rule.

The menstrual-cycle note is not a side issue. A one-year prospective cohort found perceived fluid retention peaked on the first day of flow rather than neatly premenstrually, which is exactly the kind of pattern that can make a good deficit look broken for a week.¹⁴ Some lifters also see late-luteal scale increases from higher body temperature, altered bowel rhythm, sodium sensitivity, carbohydrate intake, and training inflammation. The answer is not a special calorie target for that week. It is comparing follicular-to-follicular or luteal-to-luteal trends when the scale is noisy, and refusing to cut calories because of a predictable water phase.

The decision rule. Compute a 14-day average intake and a 14-day weight trend. Convert the weight trend to implied energy balance with a fat-energy equivalent, usually 3,500 kcal per pound or 7,700 kcal per kg. This does not revive the old static rule Hall criticized earlier. Hall's critique was about forecasting long-term weight loss as if expenditure never adapted. In calibration, the number is only a short-window translation layer: "the trend implies about this much net energy imbalance." It is a rough converter, not a promise that every 3,500 kcal deficit produces exactly one pound of tissue loss. If the implied balance matches your planned phase target within reasonable noise, hold. If it does not, change one variable at a time and run another 14 days.

A worked example. An 80 kg active man targets a 0.5 percent weekly loss, which is a 2,800 kcal weekly deficit, or 400 kcal per day. Calculator-based maintenance is 2,750 kcal per day. He sets intake at 2,350 and runs 14 days. Average weight goes from 80.0 to 79.7, a loss of 0.3 kg in 14 days, or 0.19 percent per week.

The calculator was high. The user's true maintenance was lower than the equation predicted, probably from a combination of NEAT below the assumed activity multiplier and modest under-reporting. The trend told the truth. No watch number, recipe label, or activity multiplier could have produced that conclusion as cleanly.

The same logic runs in reverse on a surplus. Suppose a 75 kg lifter has a calibrated maintenance of 2,700 kcal and wants a lean-gain phase at roughly 0.25 percent body weight per week. That is about 0.19 kg per week, which translates to roughly 1,450 kcal per week, or about 200 kcal per day. He sets intake at 2,900 kcal and holds training volume stable for 14 days.

The goal is not to make lean gain mathematically sterile. Some glycogen and gut-content increase is expected early in a surplus, and better training can make the scale look jumpy. The signal that matters is the second and third week slope. If the trend keeps running twice as fast as the target while waist and appetite are both climbing, the surplus is larger than the program needs.

09Where each tool actually belongs

Active people accumulate tools faster than they accumulate skill at using them. A wearable, a food log, a smart scale, a smart ring, and a coaching app can all coexist if each one is doing the job its data structure supports and not the job its marketing claims.

The Apple Watch-Based Calorie Targets approach, and the related How to Use Watch Calories Without Eating Back Every Workout guide, both rest on this division of labor. Use the watch to see variation. Use the trend to set intake.

10Your 21-day calibration checklist

This is where most articles end with a recap. The point of this one is to give you a calibration window you can run on yourself starting now, with a clean exit at day 21 either way.

Two specific failure modes to watch for. If your weight is moving faster than 1.0 percent per week and you have not changed intake, you are probably losing more than fat, and at least some of the loss is glycogen and water in week one or accumulated muscle loss by week six. Add 100 to 150 kcal back, hold 14 days, and recheck. If your weight is flat for three full weeks at honest logging and stable steps, the deficit is gone. The choices are a smaller intake target, a step count target, or a short maintenance phase to reset the system before another cut block. Reverse Dieting After Fat Loss covers the maintenance-block path. The First 12 Weeks of a Men's Cut covers the structured-cut path.

The model in this article is meant as a way to read your own data when something stops working. The first law of thermodynamics holds. The accounting around it is what gets noisy in practice, and the calibration loop is what keeps the accounting honest.