Why I Built a Weight Prediction Method
During my diets, I noticed that the number on the scale did not always reflect my progress clearly. My weight could remain almost unchanged for several days or even move in the opposite direction before eventually catching up with the progress I expected.
Daily body weight can fluctuate because of factors such as body water and digestive contents. This makes it difficult to determine how much weight I have actually lost simply by comparing individual measurements on the scale.
Since I was already recording my daily calorie intake, body weight, and physical activity, I wanted a second way to follow my progress. Instead of trying to predict the exact number I would see on the scale the next morning, I developed a method for estimating my underlying weight trajectory from my calorie intake and energy expenditure.
This gives me two ways to follow a diet: my measured weight shows what the scale says each day, while my predicted weight provides an estimate of the longer-term weight-loss trajectory behind those daily fluctuations.
The Idea Behind the Method
The basic idea behind my method is energy balance. My body uses energy throughout the day, while the food and drinks I consume provide energy. When the calories I burn are greater than the calories I consume, the difference creates an estimated calorie deficit that can be converted into an expected change in body weight.
For my calculations, I divide daily energy expenditure into two main parts: my basal metabolic rate (BMR), which estimates the energy my body uses at rest, and the calories I burn through physical activity. I then compare their total with my daily calorie intake.
In its simplest form, the method follows this relationship:
Daily calorie deficit = BMR + activity calories − calorie intake
I then convert this estimated calorie deficit into a predicted daily weight change. By accumulating these small changes from one day to the next, I can build a predicted weight trajectory and compare it with my measured weight over the course of a diet.
Calculating Basal Metabolic Rate
The first part of my estimated daily energy expenditure is basal metabolic rate (BMR), which estimates how much energy the body uses at rest to maintain basic functions.
For my calculations, I use the Mifflin–St Jeor equations:
| Men’s BMR | 10 × weight (kg) + 6.25 × height (cm) − 5 × age (years) + 5 |
| Women’s BMR | 10 × weight (kg) + 6.25 × height (cm) − 5 × age (years) − 161 |
For example, for someone weighing 76 kg (167.6 lb) , with a height of 175 cm and an age of 30, the estimated BMR is approximately 1,709 calories per day for a man and 1,543 calories per day for a woman.
An important part of my method is that I do not keep BMR constant throughout a diet. The calculation uses the recorded weight for each individual day. As weight changes, the estimated BMR changes with it, allowing the model to continuously adjust rather than relying on a single BMR calculated at the beginning of the diet.
Estimating Calories Burned From Walking
Walking has been a major source of physical activity during my diets, so the next part of the method estimates the calories burned from my daily steps.
I use the standard MET-based calorie calculation:
Calories burned per minute = MET × 3.5 × weight (kg) / 200
For my regular walking pace of approximately 10 min/km (16:06 min/mile), I use a MET value of 4.8. I also typically walk at around 120 steps per minute, which allows me to estimate walking time directly from my daily step count:
Walking time (minutes) = number of steps / 120
Combining the two gives:
Walking calories = 4.8 × 3.5 × weight / 200 × steps / 120
For example, at a body weight of 76 kg (167.6 lb), 12,000 steps correspond to approximately 100 minutes of walking and an estimated 638 calories burned.
As with BMR, the calculation uses the recorded body weight for each day. This means the estimated calories burned from the same number of steps gradually change as body weight changes during the diet.
Adjusting for Running and Other Activities
The walking calculation works well when most of my daily steps come from walking at my usual pace. However, not every step represents the same amount of energy expenditure. Running, hiking, and other activities can burn substantially different amounts of energy even when they contribute steps to the same daily step count.
For example, consider a 10 km (6.2 mile) run at a pace of 5 min/km (8:03 min/mile), with a running cadence of 180 steps per minute and a body weight of 76 kg (167.6 lb). The run takes 50 minutes and contributes approximately 9,000 steps:
Run steps = 50 × 180 = 9,000 steps
Using a MET value of 11.8 for running at approximately this pace, the estimated energy expenditure is about 785 calories.
My regular walking formula would already count those 9,000 steps as 75 minutes of walking at 120 steps per minute. Using the 4.8 MET walking estimate, those steps would contribute approximately 479 calories to the model.
I therefore calculate the difference:
Activity adjustment = actual activity calories − walking calories already counted
Activity adjustment = 785 − 479 = 306 calories
In this example, I would add approximately 306 calories to that day’s estimated energy expenditure. This avoids counting the run twice: the normal step calculation remains in place, while the activity adjustment accounts only for the additional energy expenditure that the walking-based model did not capture.
The same idea can be applied to other activities whenever their energy expenditure differs substantially from what my normal step-based walking calculation would estimate.
My Weight Prediction Formula
After estimating BMR, walking expenditure, and any necessary activity adjustment, I combine them with my daily calorie intake to estimate my daily weight change.
The general formula I use is:
Predicted daily weight change (g) = (BMR + walking calories + activity adjustment − calorie intake) / 7.709
The value 7.709 comes from using approximately 7,709 calories as the energy equivalent of one kilogram of body fat. Dividing by 7.709 gives the predicted change directly in grams.
For my regular walking days, the formula can be expanded as:
Predicted daily weight change (g) =
[BMR + (4.8 × 3.5 × weight / 200 × steps / 120) − calorie intake] / 7.709
If I have additional activity that is not accurately represented by my normal walking calculation, I include the activity adjustment described in the previous section.
A positive result represents predicted weight loss, while a negative result represents predicted weight gain. I calculate this value for each day and use it to update my predicted weight over time.
A Step-by-Step Example
The easiest way to see how the method works is to apply it to a complete day. The following examples use a body weight of 76 kg (167.6 lb), a height of 175 cm, an age of 30, and a daily calorie intake of 1,800 calories. The first example represents a regular walking day, while the second shows how the calculation changes when some of the daily steps come from running.
A Regular Walking Day
Suppose I record 12,000 steps during a day, with those steps coming primarily from walking at my usual pace of approximately 10 min/km (16:06 min/mile).
First, I calculate BMR:
BMR = 10 × 76 + 6.25 × 175 − 5 × 30 + 5
BMR ≈ 1,709 calories
At a cadence of approximately 120 steps per minute, 12,000 steps represent about 100 minutes of walking:
Walking time = 12,000 / 120 = 100 minutes
Using the walking calculation from earlier:
Walking calories = 4.8 × 3.5 × 76 / 200 × 100
Walking calories ≈ 638 calories
The estimated calorie deficit for the day is therefore:
Calorie deficit = 1,709 + 638 − 1,800 = 547 calories
Finally, I convert the calorie deficit into predicted weight change:
Predicted weight loss = 547 / 7.709 ≈ 71 grams
According to the model, this day would therefore contribute approximately 71 grams of predicted weight loss to the cumulative weight trajectory.
A Day With Additional Exercise
Now suppose I run 10 km (6.2 miles) at 5 min/km (8:03 min/mile), with a cadence of approximately 180 steps per minute. The 50-minute run contributes about 9,000 steps.
The model initially treats those 9,000 steps as:
Walking time = 9,000 / 120 = 75 minutes
At 76 kg, the normal walking calculation assigns approximately 479 calories to those steps. However, the 10K run itself is estimated to burn approximately 785 calories.
The required activity adjustment is therefore:
Activity adjustment = 785 − 479 = 306 calories
Using the same BMR and an intake of 1,800 calories, the complete calculation becomes:
Calorie deficit = 1,709 + 479 + 306 − 1,800 = 694 calories
The predicted weight change is:
Predicted weight loss = 694 / 7.709 ≈ 90 grams
The important point is that I do not add the entire 785 calories from running on top of the step calculation. The 9,000 steps have already contributed approximately 479 calories to the model, so I add only the 306-calorie difference. This allows the same basic formula to account for activities that differ from my regular walking pattern without double-counting them.
From Daily Changes to Predicted Weight
Calculating the predicted weight change for a single day is only the first step. To follow my progress throughout an entire diet, I accumulate these daily changes to create a predicted weight trajectory.
I begin with my measured weight at the start of the diet. After calculating the predicted weight loss for each day, I subtract it from the previous predicted weight:
New predicted weight = previous predicted weight − predicted daily weight loss
For example, if my predicted weight is 76.000 kg and the model calculates a 71-gram loss for that day, the next predicted weight becomes:
76.000 − 0.071 = 75.929 kg
I repeat this calculation each day using that day’s weight, calorie intake, steps, and any necessary activity adjustment. The result is a continuous predicted weight trajectory that I can compare with my measured weight throughout the diet.
Why Predicted and Measured Weight Can Differ
My predicted weight is not intended to match the number on the scale every day. Measured body weight can fluctuate because of changes in body water, glycogen, food and digestive contents, sodium intake, and other short-term factors that are not directly represented by the prediction formula.
Because of these fluctuations, measured weight may remain above or below the predicted weight for several days or even longer before moving back toward it. A temporary difference between the two therefore does not necessarily mean that the underlying weight-loss trend has changed.
For this reason, I focus more on how the predicted and measured weights behave over longer periods than on their difference on any individual day. In my diet data, I can compare the two trajectories over time and observe whether temporary differences eventually decrease as the diet progresses.
The goal of the model is therefore not to predict tomorrow morning’s exact scale reading, but to estimate the underlying direction and progression of my weight over the course of a diet.
Assumptions and Limitations
My weight prediction method is an approximation rather than an exact physiological model. Each part of the calculation introduces some uncertainty, including estimates of BMR, calories burned during physical activity, calorie intake, and the conversion between calorie deficit and expected weight change.
The method uses approximately 7,709 calories as the energy equivalent of one kilogram of body fat. In reality, changes in body weight do not come exclusively from body fat, so this conversion should be treated as a practical approximation rather than an exact biological relationship.
The model also estimates my daily expenditure primarily from BMR and recorded physical activity. It does not separately calculate every source of energy expenditure throughout the day, and activities measured through steps or MET values are themselves estimates.
For these reasons, I use the method to estimate my longer-term weight trajectory rather than to predict an exact scale measurement. It is a personal quantitative method that I test against my own diet data, not a medically validated weight-loss model or a substitute for professional health advice.
Testing the Method
A prediction method becomes much more useful when its results can be compared with real data. For each diet period where I have sufficiently complete and verified records, I apply the same prediction method to my daily calorie intake, activity, and weight data and compare the resulting predicted trajectory with my measured weight.
I currently have four verified datasets from my diets in 2021, 2022, 2024, and 2026. Rather than evaluating the method based on a few individual days, I examine how the predicted and measured weights behave throughout each complete diet period, including periods when they temporarily move apart and later come closer together.
The detailed data for each period are available on their individual Diet pages, while the Prediction Model Results page brings the results together to evaluate how the method has performed across the datasets tested so far.