Intro
Fueling and recovery are central to performance in elite football. Both depend on one critical variable: a player’s actual energy expenditure (EE). Even moderate errors in EE estimation can compromise glycogen resynthesis, recovery, and long-term health.
The challenge is that widely used field methods do not align with gold-standard measures. Validation studies report systematic errors. For example, GPS-based metabolic power has underestimated EE by ~19% overall, up to ~44% at high intensities, and paradoxically overestimated low-speed walking by ~43%⁵. Other work found anaerobic cost overestimated by ~300% and aerobic cost underestimated by ~40% in intermittent protocols⁶.
Because energy availability (EA) = (EI − EEE) / FFM, such errors have practical consequences. International consensus statements emphasize that EA calculations in applied sport are inherently difficult and prone to error, largely because EE is so challenging to measure reliably⁷ ⁸.
At CarboPlanner, we are working to close this gap with football-specific, multi-sensor machine learning models.
Why Precision in Energy Estimation Matters in Football
Football is highly intermittent: sprinting, accelerations, decelerations, and pauses constantly alternate. As a result, EE is not steady but fluctuates with match context, playing position, and period of play.
Studies illustrate this variability. A pilot study in professional female players suggested that EE during matches averaged ≈452 kcal/hour, compared to ≈353 kcal/hour in training³. While limited in scale, it highlights how match demands exceed training loads.
Without precise monitoring, players risk under-fueling (reducing performance and recovery) or over-fueling (impacting body composition). For clubs, moving from rough estimates to accurate profiles can unlock a competitive edge.
Limitations of Current Methods
Metabolic Power
The metabolic power model, common in GPS systems, estimates EE from speed and acceleration. While widely adopted, validation studies show clear limitations under football-specific conditions:
-
Brown et al. (2016): systematic underestimation by ~19% overall, ~44% in high-intensity drills, and ~43% overestimation at low speeds⁵.
-
Brochhagen et al. (2025): reported anaerobic energy cost overestimated by ~300% and aerobic underestimated by ~40% in intermittent protocols⁶.
Recent refinements have improved accuracy. An updated algorithm tested in Serie A players (Savoia et al., 2020) showed negligible bias compared to indirect calorimetry⁹. Still, even improved MP mainly captures locomotor load. It misses costs of collisions, jumps, technical actions, and fatigue dynamics¹⁰.
Linear Calibration Approaches
Other methods rely on calibrations, such as mapping heart rate to VO₂ from lab tests. These may work in steady-state running but struggle under intermittent football demands:
-
da Silva et al. (2021): HR-based EE estimates deviated significantly from VO₂ measures in futsal matches¹¹.
-
Gastin et al. (2018): wearable EE estimates showed large errors in intermittent, high-intensity movements¹².
Summary: Standard methods have contributed to monitoring but still fall short of the precision required in elite football.
Our Approach: Data + Machine Learning
CarboPlanner is developing a machine learning model built for football, integrating data from GPS, accelerometers, heart-rate sensors, and portable oxygen monitoring.
Rather than relying on a single input, multi-sensor fusion captures the full energetic profile of football—distinguishing locomotor from non-locomotor work, modeling both aerobic and anaerobic contributions, and accounting for individual variability.
This builds on a growing scientific consensus:
-
Tikkanen et al. (2014): no single sensor is consistently accurate—HR works best at high intensity, accelerometers at steady running, EMG at low-intensity work—suggesting that combinations outperform single inputs¹³.
-
Savoia et al. (2020): refinements to EE estimation models can markedly reduce bias compared to VO₂⁹.
CarboPlanner extends this by applying machine learning to football-specific datasets, aiming for validated, individualized EE estimates that practitioners can trust.
What It Could Mean in Practice
-
For nutritionists and performance staff: fewer assumptions and stronger data-assisted foundations for fueling strategies.
-
For players: greater trust in nutrition plans and recovery routines.
-
For clubs: scalable, automated precision nutrition that replaces generic templates with tailored insights.
Conclusion
Traditional approaches like metabolic power and linear calibration have advanced performance monitoring, but they cannot yet deliver the accuracy modern football requires.
CarboPlanner is working on the next step: football-specific, validated EE models that replace guesswork with science. As publications of our results emerge, we aim to help clubs turn energy management into a measurable performance advantage.
👉 Want to see how precise energy estimates could work in your club? Book a demo with CarboPlanner and explore the future of data-assisted fueling.
References
¹ Dasa M. et al. (2022). Accuracy of Tracking Devices’ Ability to Assess Exercise Energy Expenditure in Professional Female Soccer Players. Int J Environ Res Public Health, 19(8):4770. https://doi.org/10.3390/ijerph19084770
² Mountjoy M. et al. (2018). IOC consensus statement on relative energy deficiency in sport (RED-S). Br J Sports Med, 52:687–697. https://doi.org/10.1136/bjsports-2018-099193
³ Dobrowolski H., Włodarek D. (2023). Energy expenditure during training and matches in professional female soccer players. Rocz Panstw Zakl Hig, 74(2):143–150. https://pubmed.ncbi.nlm.nih.gov/37309847/
⁴ Anderson L. et al. (2017). Daily energy expenditure and energy intake of professional soccer players. Appl Physiol Nutr Metab, 42(10):1105–1111. https://doi.org/10.1139/apnm-2017-0048
⁵ Brown D.M. et al. (2016). Metabolic power method: underestimation of EE in field-sport movements. Int J Sports Physiol Perform, 11(8):1067–1073. https://doi.org/10.1123/ijspp.2016-0021
⁶ Brochhagen J. et al. (2025). Validity of the metabolic power model in intermittent running. Front Sports Act Living. https://doi.org/10.3389/fspor.2025.XXXX
⁷ Burke L.M. et al. (2018). Pitfalls of Conducting and Interpreting Estimates of Energy Availability in Free-Living Athletes. Int J Sport Nutr Exerc Metab, 28(4):350–363. https://doi.org/10.1123/ijsnem.2018-0142
⁸ Areta J.L., Taylor H.L., Koehler K. (2021). Low energy availability: history, definition and evidence of its endocrine, metabolic and physiological effects in humans. Eur J Appl Physiol, 121(1):1–21. https://doi.org/10.1007/s00421-020-04516-0
⁹ Savoia C. et al. (2020). Validation of an updated metabolic power approach in elite soccer players. Int J Environ Res Public Health, 17(9):3285. https://doi.org/10.3390/ijerph17093285
¹⁰ Osgnach C. et al. (2010). Energy cost and metabolic power in elite soccer. Med Sci Sports Exerc, 42(1):170–178. https://doi.org/10.1249/MSS.0b013e3181ae5cfd
¹¹ da Silva H.S. et al. (2021). Validity of HR methods to estimate EE in futsal. Front Psychol, 12:711107. https://doi.org/10.3389/fpsyg.2021.711107
¹² Gastin P.B. et al. (2018). Validity of energy expenditure estimates from wearable devices in intermittent sport. J Sci Med Sport, 21(3):291–295. https://doi.org/10.1016/j.jsams.2017.06.003
¹³ Tikkanen O. et al. (2014). EMG, HR, and accelerometer methods for estimating EE. Med Sci Sports Exerc, 46(9):1831–1839. https://doi.org/10.1249/MSS.0000000000000308
