Weight Prediction App: How Bayesian Modeling Forecasts Your Weight

Every calorie calculator on the internet uses the same formula: estimate your TDEE, subtract 500 calories, lose a pound a week. The problem? Your body isn’t a formula. Metabolic rate varies by 20–30% between individuals of the same height and weight. Generic calculators can’t account for your specific metabolism, adaptation over time, or daily variation in expenditure.

Sam Gym takes a different approach. Instead of starting with a generic formula and hoping it’s close, it starts with your actual data and learns your metabolic rate from what actually happens.

🧮 How the Prediction Engine Works

Sam Gym’s weight prediction uses Bayesian adaptive modeling. Here’s what that means in plain English:

1. It collects your real data. Every day you log meals (calories in) and either WHOOP or Apple Health provides calories burned (calories out). Sam Gym also records your daily weigh-ins.
2. It calculates your actual energy balance. For each complete day (where both intake and expenditure are logged), it knows your true calorie surplus or deficit.
3. It compares predicted vs. actual weight change. If the standard 3,500 calories = 1 pound formula says you should have lost 0.5 lbs but you only lost 0.2 lbs, the model adjusts.
4. It updates its belief about your metabolism. This is the Bayesian part. The model maintains a probability distribution of your metabolic rate and updates it with each new data point. More data = tighter estimates.
5. It projects forward. Using your learned metabolic rate and your recent intake/expenditure patterns, it forecasts your weight at 1, 2, 3, 4, 5, and 6 weeks out.

🔬 What Makes This Different from TDEE Calculators

Standard TDEE calculators ask for your height, weight, age, sex, and “activity level” (sedentary, lightly active, active, very active). Then they spit out a number. Problems:

• Activity level is a guess. What does “moderately active” mean? Three gym sessions? A desk job with weekend hiking? Everyone interprets it differently.
• No adaptation. Your metabolic rate changes as you lose weight, gain muscle, change activity patterns, or experience stress. Static calculators don’t adapt.
• Population averages. The Mifflin-St Jeor and Harris-Benedict equations are averages across thousands of people. You are not the average.

Sam Gym’s model doesn’t ask you to guess your activity level. It measures it (via WHOOP or Apple Health). And it doesn’t use population averages for long—within a week or two of data, it’s learning your specific metabolic response.

📈 The 6-Week Forecast

Sam Gym’s Weight tab shows six forecast boxes in a 3x2 grid, projecting your weight at weeks 1 through 6. Each projection is based on:

• Your current weight (7-day rolling average)
• Your recent calorie balance (average net calories over complete days)
• Your learned metabolic rate (Bayesian estimate)

The forecast updates daily as new data comes in. If you increase your calorie intake, the projections shift upward. If you add more training, they shift down. It’s a living model, not a static prediction.

✅ Data Quality Matters

The prediction engine has guardrails to ensure accuracy:

• Complete days only. A day only counts if both calories consumed and calories burned are logged. If you skip meal logging for a day, that day is excluded from calculations.
• Today excluded. Today’s data is always partial (you haven’t finished eating yet), so it’s excluded from all averages and predictions.
• Minimum data threshold. The model needs at least a few days of complete data before making predictions. More data = more confident forecasts.
• Outlier dampening. A single 5,000-calorie day doesn’t throw off the entire model. The rolling average approach naturally dampens outliers.

📊 The Weight Projection Chart

Beyond the 6-week boxes, the Weight tab includes a full projection chart that visualizes:

• Your historical weight data points
• The 7-day rolling average trend line
• The projected path forward based on current patterns
• Your target weight (if set) as a reference line

Below the chart, a net calories chart shows your daily energy balance over time, so you can see exactly why the weight trend is moving in a given direction.

🎯 Why Prediction Beats Goal-Setting

Most fitness apps let you set a goal weight and a target date, then calculate a daily calorie budget. The problem: life doesn’t follow a straight line. You have high-calorie weekends, low-appetite sick days, training blocks that increase hunger.

Sam Gym’s approach is different. Instead of telling you what you should eat to hit a goal, it shows you where you’re actually heading based on what you’re doing. This is more honest and more useful:

• If the forecast shows you’re trending toward your goal, keep doing what you’re doing.
• If it shows you’re stalling, you can see exactly why (calorie balance, activity level) and adjust.
• If you had a bad week, the forecast shows the real impact—which is usually less dramatic than you think.

⌚ Works with WHOOP and Apple Health

The prediction engine works with either data source for calorie expenditure:

• WHOOP: Heart-rate-derived calorie burn, the most accurate wearable measurement available. Syncs automatically via API.
• Apple Health: Aggregates data from Apple Watch, iPhone motion sensors, and third-party apps. Good coverage for most users.

Both provide the “calories out” side of the equation. Sam Gym’s AI meal logging provides the “calories in” side. The prediction engine connects them.