Raw or Cooked: How to Count Calories in Food You Make Yourself

Weigh a chicken breast before you cook it and again after, and you have two different numbers for the same dinner. Both are right. Only one of them matches the database entry you are about to tap.

This is one of the least discussed sources of error in a food log, and unlike most of them it is completely fixable. It is not a limitation of the science or of anybody's app. It is a mismatch between the state your food was in when it went on the scale and the state the entry assumes it was in.

The same chicken breast, two different numbers

USDA FoodData Central lists skinless chicken breast at 120 calories per 100 g raw and 165 calories per 100 g roasted.² Same bird, same cut, no sauce, no oil. The roasted figure is 38% higher.

Nothing was added. The USDA cooking yield tables put a roasted chicken breast at about 72% of the weight it went in at, with individual results between 61% and 82%.¹ So a 200 g raw breast leaves the oven at roughly 145 g:

One dinner, one answer, reached two ways. The trouble starts when the weight and the entry come from different states. Weigh the cooked breast at 145 g and tap the raw entry and you log 174 calories, 66 short. Weigh it raw at 200 g and tap the roasted entry and you log 330, ninety over. Neither entry is wrong and neither weight is wrong. The pairing is.

That 72% figure comes from breast with the skin on, which is not what most people cook, so it is worth checking a second way. Protein does not leave meat during roasting. FoodData Central puts skinless breast at 22.5 g of protein per 100 g raw and 31.0 g per 100 g roasted.² For the protein to be conserved, 100 g raw has to come out at 22.5 divided by 31.0, which is 73 g. The two routes agree to within a percentage point, which is about as much agreement as this subject ever offers.

Cooking removes water, not calories

Food energy comes from protein, carbohydrate, fat and alcohol. Water is not on the list and never has been.³ So when heat drives moisture out of a chicken breast, the calories do not go with it. They stay behind in a smaller parcel, and the figure per 100 g rises because the denominator shrank.

You can watch it happen in the composition data. Raw skinless breast is 73.9 g of water per 100 g; roasted it is 65.3 g.² The entire 38% gap between those two entries is evaporation.

This is why lean, wet foods change the most and why a cooked entry always looks worse than its raw twin. Across the 173 entries in the USDA yield tables, cooking yields run from 29% for microwaved bacon to 96% for a water-added cured ham, with a median of 76%.¹ Not one entry in the table gains weight.

When cooking does remove calories: the fat in the pan

The USDA tables are precise about the one real exception. They split each loss into the part that was water and the part that was fat, and the two add up to the yield. A pan-broiled patty made from ground beef in the 12% to 22% fat band comes out at 73% of its raw weight: 20.8% of the raw weight leaves as water and 6.2% as fat. Go above 22% fat and the yield drops to 69%, with the fat share of the loss doubling to 12.4%.¹

That fat is calories, and it ends up in the pan rather than on your plate. Take 100 g of raw 80/20 ground beef, which FoodData Central puts at 254 calories. At a 73% yield it comes off the heat at 73 g, and the cooked patty runs 246 calories per 100 g, so what is left on the plate is about 180 calories.² Around 74 calories stayed behind. Run the same arithmetic on 93/7, which is 152 calories raw and 182 cooked at a 77% yield, and the loss is about 12 calories. Browning and draining is worth six times as much on the fatty grade as on the lean.

Which entry is honest here depends entirely on where that fat went. Pour it off and the cooked entry is the one that reflects your dinner. Stir the beef into a bolognese and leave the fat in the sauce, which is what most of us actually do, and the fat is still on the plate: weigh it raw and use the raw entry.

Rice and pasta go the other way

Meat loses water. Grains absorb it, and the same arithmetic runs backwards with a much bigger multiplier.

Dry white long-grain rice is 365 calories per 100 g and holds 11.6 g of water. Cooked, it is 130 calories per 100 g and holds 68.4 g.² Dry pasta is 371 and cooked pasta is 158. So 100 g of dry rice is 365 calories no matter what you do to it, but by the time it reaches the plate those 365 calories are spread across roughly 280 g.

In sheer size this is the worst mistake in the article. Serve yourself 200 g of cooked rice, search "rice, white", tap the raw entry, and you have logged 730 calories against a plate holding 260. Nearly 500 calories, in one tap, on a food many people eat several times a week.

The reverse error is quieter and therefore worse. Weigh 75 g of dry rice, log it against the cooked entry, and you record 98 calories where you ate 274. That one hides 176 calories and reads like progress.

The oil in the pan, and what nobody has measured

Olive oil is 884 calories per 100 g, and USDA's own measure puts a tablespoon at 13.5 g, so about 119 calories.² Two tablespoons into a pan for a family stir-fry is about 240 calories that no photograph shows and no cooked-food entry contains, because the entry describes the food and not your pan.

I went looking for research on how much of that oil ends up on the plate rather than staying in the pan. What exists is food-science work on deep frying: batter coatings and modified starches, aimed at making industrially fried chips absorb less. I could not find anything that measured what fraction of a domestic tablespoon transfers into a pan of fried vegetables or an omelette, and I am not going to invent a number to fill the gap.

So take the conservative assumption and log the whole tablespoon. If a third of it really did stay in the pan, you have over-logged by forty calories, which is inside the noise of everything else here. If you log none of it and you cook most nights, you are under-recording by a few hundred calories a day, and that is exactly the size of gap that makes a month of honest effort read as a weight loss plateau.

It is worth doing once. Measure your usual pour into a tablespoon instead of going straight into the pan. Most people's "a splash of oil" is closer to two tablespoons than one.

How to count calories in food you make yourself

Weigh in one state, and log in that same state. That is the entire method, and the rest of this section is what it means in a kitchen.

Raw is the better default for anything you cook on its own: meat, fish, rice, pasta, pulses. A raw weight does not depend on how long you cooked it, how hot the pan ran or how much moisture you drove off. Two people roasting identical 200 g breasts can finish at 145 g and 160 g, and the USDA's 61% to 82% spread says that is normal rather than sloppy.¹ Both of them started at 200.

Cooked is the better default when the raw weight was never available to you: a pouch of pre-cooked rice, a rotisserie chicken, a portion of last night's dinner someone put in a box. That is also the territory covered in tracking meals you didn't cook, where nobody is weighing anything and the question becomes what a plate is worth rather than what it weighs.

Two habits follow from the rule and both are quick. Read past the food name to the state when you pick an entry, because "raw" and "cooked" are separate foods in every database rather than two ways of writing one. And weigh the ingredients rather than the finished dish, because once four things have been in a pan together the only defensible total is the sum of what went in.

Cooking once and eating it four times

A pot of chili that feeds you for four days looks like the hardest case for a calorie logger. It is actually the easiest, because the ratio does the work.

Weigh every ingredient as it goes in and total it once. Weigh the finished pot. Then your bowl is worth the fraction of the pot it represents:

A 450 g bowl is 450 divided by 2,400, which is 18.75% of the pot, so 418 calories. Note what you never needed to know: how much water evaporated, or how long it simmered. The total was fixed the moment the last ingredient went in, and everything after that just moved it around.

The assumption this rests on is that the pot is evenly mixed, and it is a real assumption. A stew where the meat has settled and the liquid is on top does not divide cleanly, and neither does a tray bake where one corner got all the oil. Stir before you serve, or accept that the four portions are equal on average rather than individually.

What a photo scan can see here, and what it cannot

I build a calorie counter app that reads a meal from a photograph, so this section is where I have to be straight about what mine cannot do. MyPlate has no food database and no barcode scanner. A scan is an estimate of a plate, not a measurement of it.

A photograph sees the cooked food. It cannot tell which state you weighed in, because you did not weigh, and the oil is already inside the food by the time the picture is taken. Nothing in the image says whether you drained the beef.

The measured version of that limitation arrived in 2026. Researchers ran an AI image-based dietary assessment app against doubly labelled water, the physiological reference method, in 20 women with obesity over seven days. Against a measured total energy expenditure of 3,004 calories a day, the app underestimated intake by 25%, a bias of 817 calories a day. A 24-hour dietary recall did considerably worse at 50%. Individual-level agreement was effectively absent: an intraclass correlation of 0.00, with limits of agreement running from minus 3,707 to plus 2,073 calories a day.⁴ That is twenty participants, one population and a different app from this one, and the authors' own conclusion was that the field needs standardised clinical validation before anyone leans on these tools clinically.

Self-report is no refuge either. Pooling five validation studies against recovery biomarkers, energy intake was under-reported by an average of 28% on a food frequency questionnaire and 15% on a single 24-hour recall, with substantial variation between studies and with body mass index, education and age all predicting who under-reports.⁵ There is more on how those numbers stack up in how accurate calorie tracking apps actually are.

Photograph the meals you could not weigh, which for most people is most of them. But when you cooked it yourself and the scale is already on the counter, weighing the ingredients beats anything a camera will give you, and after the first week it is barely slower.

These errors do not all point the same way, which is worth noticing. Cooked weight against a raw entry under-counts. Raw weight against a cooked entry over-counts. Cooked rice against a dry entry over-counts enormously. Because they run in both directions they do not quietly cancel, and no blanket correction fixes them. Only matching the weight to the entry does.

Common questions

Weigh one state and stay in it

Raw for what you cook, cooked for what arrives already cooked, and the entry always matched to the weight you took. On a day that includes a bowl of rice or a drained pan, that one habit is worth more than any further refinement of your calorie target.

Do not respond to a corrected log by cutting further. If fixing the rice error drops your recorded intake by 400 calories, nothing about your dinner changed; only the number on the screen did. If you are eating in a deficit, the floor worth respecting is 1,200 calories a day for women and 1,500 for men, which is where the NIH clinical guidelines put it and where MyPlate's own targets stop.⁷

And if any of this is tipping you from useful attention into anxiety, weighing every ingredient and rechecking totals you have already checked, that is the point to stop and talk to a doctor or a registered dietitian. A log you keep most days is worth more than a perfect one you abandon: across dietary self-monitoring research it is the frequency of logging, not its precision, that tracks with results.⁶

Sources

Every figure above traces to one of these. If you find a number that doesn't match the source it claims, tell me and I'll correct it.

1. Roseland JM, Nguyen QV, Williams JR, Patterson KY, Showell BA, Pehrsson PR (2014). USDA Table of Cooking Yields for Meat and Poultry, Release 2. USDA Agricultural Research Service, Nutrient Data Laboratory. Read the USDA table →
2. US Department of Agriculture, Agricultural Research Service (2018). FoodData Central, Standard Reference Legacy Release, April 2018. USDA National Agricultural Library. Read on FoodData Central →
3. Food and Agriculture Organization of the United Nations (2003). Food energy: methods of analysis and conversion factors. FAO Food and Nutrition Paper, 77. Read the FAO report →
4. Serra M, Alceste D, Jucker N, et al. (2026). Limited validity of an AI-powered app for dietary assessment in females with obesity. npj Digital Medicine, 9(1):357. Read on PubMed →
5. Freedman LS, Commins JM, Moler JE, et al. (2014). Pooled results from 5 validation studies of dietary self-report instruments using recovery biomarkers for energy and protein intake. American Journal of Epidemiology, 180(2):172-188. Read on PubMed →
6. Harvey J, Krukowski R, Priest J, West D (2019). Log Often, Lose More: Electronic Dietary Self-Monitoring for Weight Loss. Obesity, 27(3):380-384. Read on PubMed →
7. National Heart, Lung, and Blood Institute, NIH (1998). Clinical Guidelines on the Identification, Evaluation, and Treatment of Overweight and Obesity in Adults. NIH Publication 98-4083. Read the guidelines →

The formulas and limits behind MyPlate's own numbers, with their citations, are on the health sources page.