You finish a session, your watch says 520, and your app hands you 520 extra calories to spend. Eat them and the week goes nowhere. Ignore them and you're training hard on a deficit that keeps growing. The question is fair, and the honest answer is: eat back some of it, because the number you're being shown is wrong in two separate ways that both point the same direction.
Worth knowing which two, because once you can see them the arithmetic stops being a guess.
Error one, the big one: the number is a guess
Wrist devices are genuinely good at some things — step counts and heart rate hold up well against laboratory equipment. Energy expenditure is not one of them, and it isn't close.
When consumer trackers have been tested against laboratory measurement, errors on calories burned ran from 27% to 93% across the devices evaluated — around 31% for walking and running, and about 52% for cycling, because a wrist can't see how hard you're pushing. Not one device tested came close.
Two things worth being precise about, because the internet usually isn't:
- The direction varies by brand. Some devices systematically overestimate, others systematically underestimate. "Your watch is reading high" is a sensible default, not a law.
- Self-entered exercise is worse than a watch. Choosing "aerobics, 45 minutes" from a dropdown applies a population-average energy cost to a session only you saw. That's a wider margin than anything a sensor produces.
Error two: it counts calories you were burning anyway
Smaller, and almost nobody is told about it.
Your body burns energy continuously — digesting, breathing, keeping tissue alive — whether you're on a treadmill or on the sofa. When a watch or an app reports "calories burned" for a workout, it reports the gross figure: everything you burned during that hour, including the part you'd have burned sitting still.
That resting portion is already inside your daily calorie target. Your maintenance number was built from it. Count it once in the target and again in the workout credit, and you've counted it twice.
For a typical adult the resting share of an hour is roughly 70 to 80 calories. So a session credited at 500 nets out nearer 425 on this correction alone — modest next to the first error, and it runs in the same direction.
The difference has a name: gross burn is the total for the hour, net burn is the extra you spent because you trained. Net is the only one that belongs on top of a maintenance target.
The two errors compound, and they compound one way
Discount an estimate that is often out by 30% or more, then subtract the resting calories already inside your target, and the credit you can safely spend is frequently around half of what you were shown. That is where the rule of thumb comes from — not caution for its own sake, but two corrections multiplied together.
The third thing, which isn't an error at all
There's a further gap between what a session costs and what your day ends up spending, and it isn't a measurement problem — it's you.
When people add exercise, total daily energy expenditure tends to rise by noticeably less than the sessions themselves cost. Some of the difference is behavioural: you sit down more afterwards, you take the lift, you fidget less, the dog walk gets shorter. Some of it is physiological. The pattern shows up consistently enough across exercise studies — daily expenditure rising by a fraction of what the additive arithmetic predicts — that it's worth planning around rather than arguing with.
Two details make it practical rather than depressing. The gap appears to be wider when you're dieting at the same time — exactly when you least want it. And it appears to be smaller with resistance training than with steady cardio, which is one more entry on the long list of reasons to lift while you're in a deficit.
None of this means training doesn't help. It means training is not a calorie dispenser, and treating a session as a coupon is the single fastest way to stall a deficit that was working.
So what do you actually do
Three options, in ascending order of how well they work.
Eat back roughly half. Crude, and it beats both extremes. Take whatever your watch or app tells you and spend about 50% of it. You'll be approximately right most days, which is the standard worth aiming at — every calorie number is an estimate anyway, and a consistent approximation is what makes the feedback loop work.
Eat back none of it, and set a smaller deficit. Perfectly valid if your training is light or your weeks are consistent. Fold the training into your maintenance figure by choosing a slightly higher activity level, then never think about it again. The failure mode is real, though: on a hard week you end up eating far too little, and the session quality and the adherence both go with it.
Separate the two numbers properly. Build your daily target from a baseline that excludes training — a true "if I did nothing" figure — and then add each session's net cost on top of it. Nothing is double-counted, because the baseline never contained the training in the first place, and a hard day genuinely earns more food than an easy one. This is what the TDEE calculator is for: it gives you the baseline, so you know what you're adding to.
Train because it makes you stronger, healthier and better at holding a deficit — not to buy dessert. If a session has to earn its food to feel worth doing, the session was never the problem.
The cases where the answer changes
Long endurance sessions. Above about ninety minutes, refusing to eat any of it back stops being conservative and starts being a recovery problem. Fuel long training. The deficit can come from the other six days.
Very low targets. If you're already eating near the bottom of a sensible range, the argument for eating some of it back gets stronger, not weaker — an aggressive deficit plus hard training is how people lose muscle and stall anyway. As a floor: don't plan to eat much below your own resting burn, which for most people means not going under about 1,200–1,600 calories a day. Training on top of a target that low is a reason to raise the target, not to eat back less.
Strength work. A lifting session's calorie cost is genuinely modest — it's rarely the big number the display suggests. Lift for what it does to what you keep while you lose, not for the burn. Cardio and strength do different jobs, and only one of them was ever about the calories.
The honest summary
Eat back about half of what you're shown, or stop being shown a number that needs halving. If your weight trend over three or four weeks is doing what you expected, your arithmetic is fine whichever route you took. If it's flat, the exercise credit is one of the first places to look — and it is almost never that you're eating back too little.
What Sophia does with this
She never gives you a number that needs correcting, because the correction is built in. Your daily target is a baseline that deliberately excludes training, so nothing about your workouts is already inside it. When you log a session, she adds it on top as a separate bonus — the extra you actually spent, not the hour's gross total — which is why the double-count can't happen.
Then she checks her own work. She compares what you logged against what your weight trend actually did over the following weeks, and if the two disagree, she moves the numbers rather than telling you to try harder. If your training credit has been running generous, that shows up in the trend, and she adjusts it. You never have to guess whether to eat back half.