PhD Pinboard: Making Sense of Accelerometer Metrics in Team Sports

This PhD Pinboard article by Laura Dawson explores the validity, reliability, and application of accelerometer measures in tracking athlete training load.

I'm Laura Dawson, a lecturer in sports biomechanics at St Mary’s University, Twickenham, and I am currently completing my PhD at the University of Suffolk, focusing on the validation and application of the accelerometer technology and metrics within STATSports Apex Pro GNSS units for athlete load monitoring in team sports.

The last two decades have seen an explosion in the use of wearable technology across elite and sub-elite team sports. Global Navigation Satellite System (GNSS) and inertial measurement units (IMUs) – particularly accelerometers – are now widely used to monitor training loads and movement demands in elite football, rugby, and other invasion sports, but while GNSS metrics such as distance and speed are well understood, accelerometer-based metrics remain less clear.

These metrics, like PlayerLoad™, Dynamic Stress Load (DSL), and Body Load™, are widely used to monitor load, collisions, and impacts. However, as these tools become more accessible, many practitioners are asking: What do these numbers really mean?

My PhD research set out to explore exactly that - specifically, whether accelerometer-based load metrics, like STATSports’ DSL, are valid, reliable, and useful, how they are used in practice, how well they reflect athlete load, and what role they might play in informing coaching, performance, and injury-prevention decisions.

My PhD Story

I was drawn to this research after noticing the growing reliance on accelerometer metrics in professional sport, but also the confusion among practitioners about what the numbers really meant.

In conversations with coaches and sport scientists, it became clear that while many practitioners used metrics like DSL or PlayerLoad™, very few could explain their calculations, or whether they were valid and reliable enough to base training decisions on.

GNSS technology, often referred to under the GPS umbrella, has become standard for tracking player movement: distance covered, peak speed, and positional data are staples in team sport analysis. But many of the most physically demanding movements like tackles, decelerations, and vertical jumps happen with little or no horizontal displacement, and are therefore, often missed or underreported by GNSS.

That’s where accelerometers come in.

Embedded in the same upper-back wearable devices, accelerometers can measure instantaneous trunk acceleration, offering potential insight into mechanical loading during impacts and high-intensity movements. Manufacturers have created proprietary metrics such as PlayerLoad™, Body Load™, and DSL, each combining acceleration data through their own (often opaque) algorithms to provide cumulative load values.

The problem? Most practitioners are unsure what these numbers truly represent, and there’s been limited independent research validating their use.

Surveying the Field: What are Sports Practitioners Thinking and Doing?

The first phase of my research involved a large-survey of 72 sport science, medical, and coaching practitioners using GNSS and accelerometer tech across football, rugby, hockey, and other team sports (Dawson et al., 2024a).

Figure 1 - GNSS-derived (blue) and accelerometer-derived (orange) metrics used by practitioners in the monitoring of athlete training load (n=72).

In the survey, I found strong support for GNSS technology in team sports. Every respondent endorsed its use, and 70.8% felt it was important for success. This showed me how central GNSS has become in modern practice.

When we looked at metrics, GNSS-derived measures dominated (93% used total distance), while accelerometer metrics like PlayerLoad or Body Load were reported by 39% (Figure 1). For me, this highlighted a continued reliance on GNSS outputs over accelerometer data.

I also saw a clear gap in confidence. While 84.7% agreed GNSS metrics were easy to understand, only 47.2% felt the same about accelerometer-derived measures. This mismatch suggests knowledge and application remain limited.

Practitioners flagged three key needs: better education (90.3%), clearer standardisation (93.1%), and greater accessibility (70.8%), with cost a major barrier outside elite sport. Open responses often mentioned confusion around interpreting accelerometer metrics and uncertainty over thresholds. One comment summed it up: “We as a staff find it hard to really understand these metrics or exert any influence upon them in order to prescribe certain training loads.”

Despite challenges, I felt it was encouraging that practitioners still valued accelerometer measures, especially for capturing hidden loads in non-linear, high-impact, or indoor environments where GNSS is less effective.

While 84.7% agreed GNSS metrics were easy to understand, only 47.2% felt the same about accelerometer-derived measures.

Assessing Accelerometer Validity: What Does the Literature Say?

One of the key areas of focus for my PhD was establishing the validity of the accelerometer-derived that accumulate accelerations and express them as arbitrary units e.g. PlayerLoad, DSL, and Body Load. This began by evaluating convergent validity from existing literature.

When I reviewed the literature (Dawson et al., 2024b), I saw clear evidence that these accelerometer-based metrics can reflect training demands. They tended to correlate well with things we already use, such as running distance, heart rate, and even collision counts in rugby and football (Figure 2).

That gave me confidence that these metrics are valid indicators of certain aspects of load.

Figure 2. Magnitude of correlations found in existing literature for PlayerLoad, Body Load and Dynamic Stress Load against common measures of athlete load.

At the same time, I found limitations. The validity of these metrics really depends on the device, how the data is processed, and the type of activity. Raw accelerometer signals were often noisy or inaccurate, but filtering improved things a lot. For example, using a 10 Hz filter for low-impact tasks or 20 Hz for high-impact activities made the data much closer to gold-standard motion capture. This means practitioners need to be mindful of how the data is collected and processed before drawing conclusions.

I also found emerging links between accelerometer metrics, fatigue, and injury (as Michael Gerhardy also discussed in his PhD Pinboard article). PlayerLoad™, for instance, tends to rise when fatigue alters movement quality, while both high and unusually low values have been associated with greater injury risk. These findings are promising but not yet conclusive as there’s still a lot to learn about the underlying causal pathways.

Overall, I came away convinced that accelerometer-based metrics can be valuable tools, especially for collision sports or non-linear, high-intensity activities where GNSS falls short. But they’re not perfect measures of biomechanical load, and their meaning is context-dependent. The big challenge now is to better connect these external load measures to what’s happening inside the body—at the level of muscles, tendons, and bones—so we can really understand how training affects athletes over the long term.

Interunit reliability

One of the most practical concerns for load monitoring is device-to-device consistency. If different GNSS units worn by athletes produce different values for the same movement, it undermines trust in the data especially when tracking small changes over time.

Figure 3 – 2 x 20 m, 4 x 10 m and 8 x 5 m repeated sprint change of direction protocols

Figure 4 - STATSports Apex Pro GNSS unit alignment in harnesses on the upper-back for assessment of interunit reliability.

Chapter 5 of my thesis (Dawson et al., 2025) investigated the inter-unit reliability of the STATSports Apex Pro GNSS accelerometers. I worked with 33 female football players, who completed three shuttle protocols: 2 × 20 m, 4 × 10 m, and 8 × 5 m.

Each player wore two devices at the same time, allowing me to compare unit-to-unit agreement for both GNSS-derived metrics (distance, speed, accelerations, decelerations, metabolic power) and accelerometer-derived metrics (DSL and Fatigue Index: Fatigue Index = Dynamic Stress Load_i / Speed_i^k, where ‘k’ is a weighting factor: Beato et al., 2021).

The results were very clear. GNSS metrics showed good to excellent reliability, with values like total distance, maximum speed, and speed intensity almost identical between devices (ICC values above 0.9 and percentage bias close to zero). This means practitioners can be confident that these metrics are consistent across units.

However, GNSS-derived acceleration and deceleration counts were less consistent than distance or speed. Although they reached “good” reliability overall, the variability was higher, particularly in the shorter, high-turnover shuttles. This suggests these metrics should be interpreted more cautiously than total distance or maximum speed.

For the accelerometer-based measures, both DSL and the Fatigue Index performed poorly, with ICCs around 0.48–0.50 and very wide limits of agreement between devices. In other words, the same athlete could record noticeably different values depending on which unit they wore. This issue became worse as the shuttle runs included more changes of direction, which likely added extra noise into the accelerometer signals.

ICC interpretation from Koo et al., 2016

For me, the big takeaway is that GNSS-based distance and speed metrics are trustworthy across units, but accelerometer-derived metrics remain a problem. Until the algorithms and hardware improve, practitioners should avoid swapping devices between players or across sessions if they want consistent data. Assigning the same unit to each athlete is the safest approach.

This study reinforced what I’d seen earlier in my review: accelerometers have real potential to capture hidden biomechanical loads, but right now, their reliability lags behind. For applied staff, that means sticking with GNSS-derived measures for confident decision-making, while treating accelerometer outputs as exploratory rather than definitive.

Final Thoughts

Accelerometer metrics have the potential to add value by capturing movement demands that GNSS alone cannot detect, such as:

• Vertical efforts (jumps, landings)

• High-frequency changes of direction

• Impacts and collisions

• Fatigue-induced movement variability

These capabilities make them particularly useful for collision sports, indoor environments, and rehabilitation scenarios where subtle changes in movement strategy may indicate readiness or risk.

However, as wearable technologies become more advanced, the flood of available data can feel overwhelming. My PhD findings so far suggest that accelerometer-based metrics can provide valuable insights, but only when used as part of a theory-informed, context-aware framework.

Practitioners don’t necessarily need to know every line of code behind a metric, but it's vital that practitioners understand what their specific metric is (and isn’t) measuring, use consistent devices, and interpret data with contextual caution.

Ongoing work will continue to refine our understanding of how reliable and valid these metrics truly are. But even now, with a clearer grasp of their limitations and capabilities, they can already play a meaningful role in team sport load monitoring and performance management.

FAQs on Accelerometers & GPS/GNSS

What does an accelerometer measure in sport?

An accelerometer measures changes in acceleration in three directions (up-down, side-to-side, and forward-backward). In sports wearables, this data is used to estimate whole-body load from activities like sprinting, jumping, tackling, and changing direction.

What’s the difference between GPS metrics and accelerometer metrics?

GPS (or GNSS) metrics measure movement in space—things like distance covered, speed, and position. Accelerometer metrics, on the other hand, measure the forces acting on the body from impacts, accelerations, and changes of direction.

Are GPS and GNSS the same thing?

No, GPS is one satellite system, while GNSS (Global Navigation Satellite Systems) includes multiple systems such as GPS, GLONASS, Galileo, and BeiDou. Modern athlete monitoring units typically use GNSS, which allows access to more satellites, improving accuracy and reliability. See this introduction to the technology for more.

What is PlayerLoad™?

PlayerLoad™ is a Catapult proprietary metric calculated from accelerometer data. It combines movement across three axes into a single number that reflects the overall “load” an athlete experiences from impacts, accelerations, and decelerations. PlayerLoad™ is derived from the rate of change in acceleration in three planes (x, y, z) measured by the device’s accelerometer. The values are combined using a vector sum and then divided by 100 to give a simplified load number (Nicolella et al., 2018).

What is Dynamic Stress Load (DSL)?

Dynamic Stress Load is a STATSports accelerometer-based metric that sums the weighted impacts above a certain threshold (e.g., >2 g) (Beato et al., 2021). It is designed to capture the stress from high-intensity, non-linear, or contact-heavy movements.

Why are accelerometer metrics less reliable than GPS metrics?

Accelerometer signals are influenced by factors like device placement, harness movement, and the thresholds used in proprietary algorithms. This makes them noisier and less consistent between units compared to distance and speed metrics derived from GNSS.

References

Beato, M., & Drust, B. (2021). Acceleration intensity is an important contributor to the external and internal training load demands of repeated sprint exercises in soccer players. Research in Sports Medicine, 29(1), 67–76.

Dawson, Laura; McErlain-Naylor, Stuart A.; Devereux, Gavin; Beato, Marco. Practitioner Usage, Applications, and Understanding of Wearable GPS and Accelerometer Technology in Team Sports. Journal of Strength and Conditioning Research 38(7):p e373-e382, July 2024. | DOI: 10.1519/JSC.0000000000004781

Dawson, Laura; Beato, Marco; Devereux, Gavin; McErlain-Naylor, Stuart A. A Review of the Validity and Reliability of Accelerometer-Based Metrics From Upper Back–Mounted GNSS Player Tracking Systems for Athlete Training Load Monitoring. Journal of Strength and Conditioning Research 38(8):p e459-e474, August 2024. | DOI: 10.1519/JSC.0000000000004835

Dawson, L., McErlain-Naylor, S. A., Devereux, G., & Beato, M. (2025). Interunit reliability of STATSports APEX global navigation satellite system and accelerometer-derived metrics during shuttle run protocols of varied distances and change of direction frequency. Journal of Sports Sciences, 1–13. https://doi.org/10.1080/02640414.2025.2555554

Koo, T. K., & Li, M. Y. (2016). A Guideline of Selecting and Reporting Intraclass Correlation Coefficients for Reliability Research. Journal of Chiropractic Medicine, 15(2), 155–163.

Nicolella, D. P., Torres-Ronda, L., Saylor, K. J., & Schelling, X. (2018). Validity and reliability of an accelerometer-based player tracking device. PloS one, 13(2), e0191823.