Estimating Exercise Energy Expenditure: When the Wearable Number Lies

Exercise Energy Expenditure (EEE) is the most error-prone input in any sports nutrition calculation. It feeds your energy availability estimate, your daily kcal target, your periodized fueling plan, and your post-session refueling protocol. Get EEE wrong by 300 kcal a day and an LEA estimate flips from green to red — or vice versa. The athlete's wearable says 1,420 kcal. The athlete's coach says they had a "huge session." The number on your chart says something else again.

Most sports RDs I work with default to whatever the wrist-worn device reports and adjust by feel. That works until it does not — until the day a chronically underfueled athlete is showing 30 kcal/kg FFM on paper because the wearable is overstating expenditure by 25%, and the screen looks fine when it is not.

This is the framework I use to size EEE for clinical work. It is not a research protocol. It is a working method that gets EEE within a defensible band — usually ±10–15% of indirect calorimetry — without doubly-labeled water.

Why the Wearable Number Is Wrong

Wrist-worn devices estimate expenditure from heart rate plus a proprietary algorithm that infers movement type, intensity, and a personal calibration drawn from age, sex, weight, and sometimes VO2max estimates. Three structural problems:

Resistance training breaks the model. HR-based algorithms were validated against cyclical aerobic activity. A heavy back squat raises heart rate via sympathetic drive and isometric loading, not via cardiac output proportional to oxygen consumption. The wearable interprets the HR spike as continuous aerobic effort and reports a number that can be 40–80% too high. Studies comparing wrist devices to indirect calorimetry during resistance work routinely converge on overestimation in this range across the FitBit, Apple Watch, and Garmin literature.

Mixed-modality sessions average to noise. A typical team-sport practice — warm-up jog, dynamic, drill blocks, scrimmage, conditioning, cool-down — is six different activities with six different efficiency curves. The wearable picks one model for the whole session. The result is a number that is roughly directionally correct on aggregate but can be 200–500 kcal off in either direction on any given day.

Calibration is shallow. The device knows the athlete's weight (which the athlete may not have updated in months), age, and sex. It does not know FFM, sport, training history, fitness level, or whether today is a hard session or a tempo day. A 95 kg lineman and a 95 kg basketball guard with the same average HR get the same kcal number for the same session. They should not.

The device is useful. The number it reports is a starting point, not an answer.

The Three-Method Triangulation

For any clinical EEE estimate, I use three methods in parallel and reconcile them. Each is imperfect; the spread between them is diagnostic.

### Method 1: Wearable HR-Based Estimate (with Caveats)

Take the wearable's reported kcal for the session, but apply two adjustments:

• Resistance work: discount by 30–40%. If the session was pure lifting, the device number is the upper bound; trim accordingly.
• Subtract the RMR-equivalent for the session duration. A 60-minute session for an 80 kg athlete with an estimated RMR of 1,800 kcal/day burns roughly 75 kcal at rest in that same hour. Wearables that report "active calories" already do this; wearables that report "total calories" do not. Check which mode the athlete's device is in.

For aerobic-dominant sessions in fit athletes wearing chest straps, this adjusted number is usually within 10–15% of truth. For wrist-worn devices on resistance or mixed sessions, treat it as noisy.

### Method 2: MET-Based Estimate

The Compendium of Physical Activities (Ainsworth et al., maintained by Arizona State) lists METs for hundreds of activities. The formula:

EEE (kcal) = METs × body mass (kg) × duration (h) − RMR-equivalent for duration

A 75 kg athlete running at 8 mph (13.5 METs) for 45 minutes:

• Gross: 13.5 × 75 × 0.75 = 759 kcal
• Net (subtract RMR equivalent): 759 − ~70 = ~690 kcal

MET values are population averages, not athlete-specific. They are most accurate at moderate intensities and most wrong at very high intensities (where elite athletes are more efficient than the population) and very low intensities (where individual variation dominates). For a fit endurance athlete, expect MET-based estimates to overstate by 5–10%.

For resistance training, the Compendium gives roughly 6 METs for "vigorous effort" weight lifting. That is broadly fine for circuit work or hypertrophy at moderate loads. For heavy strength work with long rests (5×5 at 85%+), it overstates because the rest periods are largely passive — use the activity-specific method below.

### Method 3: Activity-Specific Heuristics

For session types where Methods 1 and 2 both struggle, I keep a reference table of activity-specific kcal-per-minute estimates triangulated against indirect calorimetry over the years. The shortlist:

• Heavy resistance training (working portion only): 5–7 kcal/min for the actual sets, plus ~1.5 kcal/min during rest. A 60-minute session with 25 minutes under load: 25 × 6 + 35 × 1.5 ≈ 200 kcal net.
• Steady-state aerobic (Z2 cycling, easy run): 8–12 kcal/min for the working portion, scaled by body mass.
• HIIT / interval running: 14–18 kcal/min during work intervals, 4–6 kcal/min during rest, summed across the session.
• Skill-based team practice (basketball, soccer scrimmage): ~9–11 kcal/min average across the active portion, scaled by mass.
• BJJ / wrestling live rolling: 12–16 kcal/min during live rounds; lower during drilling.

These ranges come from a mix of calorimetry studies, GPS-based work-rate inference, and direct reconciliation against athlete logs over the years. They are not gospel. They are sturdier than a wrist device on most resistance and mixed sessions.

Reconciling the Three Numbers

Compute all three. Look at the spread.

Spread under 15%: pick the middle estimate. Document the reconciliation in your Objective section: "EEE estimate 615 kcal — wearable 680, MET 590, activity-specific 580 — reconciled to 600."

Spread 15–30%: identify which method is most likely wrong for this session type. Resistance-heavy session and the wearable is the high outlier? Discount it. Endurance session and the wearable agrees with the MET method but the activity-specific number is low? Trust the first two. Document the choice and the reasoning.

Spread above 30%: something is broken. Either the session description is incomplete (the athlete did 90 minutes but only logged the 45-minute lift portion), the device malfunctioned (chest strap dropped), or the activity does not fit any of your reference models. Flag in the chart and re-estimate with better session data before using the number for any downstream calculation.

The discipline here is not about getting the "right" number — there is no right number without indirect calorimetry. It is about producing an estimate you can defend, with documented reasoning, that is unlikely to be wrong by 500 kcal in either direction.

Frequency and Documentation

For a flagged athlete in an LEA workup, EEE estimation runs for 7 consecutive days alongside the food log. For a routine fueling consult, a representative 3-day sample (one hard, one moderate, one rest or recovery) is adequate.

Document EEE in the Objective section of the SOAP note with three fields:

• Method or methods used
• Reconciled value (in kcal, and in kcal/kg body mass for trend tracking)
• Confidence band ("±10%" / "±20%" / "low confidence — flag")

A reader picking up the chart in three months should be able to tell whether the EEE number is a tight estimate or a rough one. That changes how aggressively the next RD can interpret an LEA flag built on it.

The Common Mistakes

Trusting the wearable number alone. Convenience is not accuracy. The number on the wrist is one input, not the answer.

Failing to subtract RMR. Most "calories burned" displays report gross expenditure during the session, including the kcal the athlete would have burned at rest in that hour anyway. EA math requires net EEE. A 75-minute session at a 200-kcal-RMR-equivalent baseline costs ~250 gross kcal of "background" that the EA formula already accounts for elsewhere in the intake-vs-resting math. Subtracting matters.

Counting warm-up and cool-down as primary work. A 90-minute "session" that is 25 minutes of working sets and 65 minutes of warm-up, mobility, and rest is not a 90-minute lift. Estimate the actual work portion and the actual rest portion separately.

Using a single method across all session types. A wearable that reports tightly on aerobic Z2 work will mislead you on resistance training. The triangulation matters most where each individual method is weakest.

Not updating body mass before the calculation. MET-based EEE is mass-dependent. A 78 kg athlete who weighed 81 kg last quarter and has not updated is generating estimates 4% too high. For monthly EA tracking, this drift compounds.

Where Platform Tooling Helps

The bottleneck is the per-session reconciliation. For a single athlete on a single day, three methods plus a reasoned pick takes 5–10 minutes. For 8 flagged athletes running 7 days each, that is 4–9 hours per cycle of pure arithmetic.

The leverage point is integrated wearable data alongside the food log, with both the device's reported kcal and a parallel MET-based or activity-specific computation pre-populated for each session — leaving the RD only the reconciliation decision and the documentation note.

That math — minutes per session reconciliation versus tens of minutes for full per-session estimation — is the difference between running a defensible EEE workflow on every flagged athlete and running it on the two who are loudest.

The Bottom Line

The wearable number is a starting point. A defensible EEE estimate is a triangulated, documented, session-specific number with a confidence band — not a screenshot from the athlete's app.

Build the three-method triangulation into your workflow. Discount wearable HR estimates for resistance work. Document the reconciliation in the Objective section. Audit your roster's EEE estimates against session type quarterly to catch drift.

If your LEA estimates are built on raw wearable numbers, your LEA estimates are noisier than you think.

The triangulated EEE feeds into the broader [member compliance audit](/blog/member-compliance-auditing-in-sports-nutrition) — the three-stream framework that reads the food log, body data, and performance data against each other every two to four weeks. Without an accurate EEE the performance-data stream is noise, and the audit collapses to a log-versus-weight comparison only.

[Calsanova's Dietitian plan](/signup?role=dietitian) integrates wearable data and food logs into a single session-level dashboard, runs the three-method EEE triangulation automatically, and surfaces the spread for RD reconciliation. Start your 30-day free trial and tighten the most error-prone number in your clinical workflow.