This section provides an overview into the use of accelerometry in sports and exercise. Steve Barrett very kindly agreed to give us his insights. I have known Steve since 2008 and he was the first person in the UK to work with the Catapult GPS system (for Perform Better UK). He has worked with this system for longer than anyone I know and therefore I feel best placed to share his thoughts on accelerometry. Steve is a very busy man combining his work with Catapult and England Women’s under 23 soccer team with his continuing doctoral research studies, hence I am very grateful to him for his time. Here’s what he has to say.


Accelerometers have recently been introduced within the elite sport domain to quantify and monitor the external training loads within different sports as a measure of mechanical load. As with all methods used currently it is important to understand the origin and the evolution of these currently used techniques to establish how they can be effectively applied in monitoring athletes. Accelerometers within pedometers measure the number of steps taken by an individual which are identified by the spikes in the uni-lateral accelerometer working in the vertical plane. In sedentary populations this methodology has helped to increase physical activity by prescribing a set number of steps/activities to perform in order to reduce obesity and cardiovascular diseases. This led accelerometers to be incorporated into new technology, where the data was originally assessed as accelerometer counts per minute. However, accelerometers were mainly applied within short and low level physical activity tasks or treadmill running (Halsey et al., 2011; Howe et al., 2009). Now we have transferred the application of this data into intermittent, multi-directional, high intensity exercises such as soccer match-play. Have we potentially skipped steps within the research in order to understand the true meaning of this data we are receiving?

Anecdotally, reports indicate that monitoring accelerometry data can potentially help the prediction of injury onset by detecting alterations in the ‘normative data output’ from an individual. Through the introduction of MEMS devices (microelectircal mechanicalsystems; such as GPS devices containing other inertial sensors such as accelerometers) it has become obvious that using accelerometer counts per minute within high intensity activities such as soccer match play are not easy to manage as a result of large unmanageable numbers. Manufacturers in conjunction with research partners introduced algorithms into their software to make this accelerometry data easier to manage (accumulating the accelerations in three planes of movement). This was the objective of the equation integrated into the Catapult™ software which was designed in conjunction with the Australian Institute of Sport for calculation of PlayerLoad™ from accelerometry data. Other systems do use alternative algorithms, however, PlayerLoad™ is the only algorithm within these MEMS devices used in the elite sport domain which has had their algorithm/ accelerometer tested for reliability and reproducibility (Boyd et al., 2012).

Interestingly, the use of accelerometry within the team sports research is still relatively new and with few research articles published on them (pubmed search for tri-axialaccelerometers in sport results in 14 findings, with player load in sport producing 55 research papers, 6 specifically investigating the accelerometry derived player load). The research articles out there are descriptive in nature and very similar to the initial papers on motion analysis within the team sport domain, where basic variables, are being reported. If we continue to look at accelerometry data as a similar accumulative measurement as distance is in sport, then we potentially don’t see the full picture for our athletes/players in training and competition.

One way that we have to view all our data is to analyse it for what it is and not become over concerned with knowing which is better. If we take heart rate, accelerometers and RPE, we are using 3 variables that are measuring three different strands to our overall loading model. Internal (heart rate) is a tool to assess how our cardiovascular system copes with the external and sometimes psychological loads placed upon the body. With RPE we are asking players how hard they found the session, which is why we see such strong correlations with heart rate and accelerometer data in certain scenarios and environments. Accelerometers (tri-axial) assess the accelerations our body go through in the different planes of movement. Again the accelerations and decelerations we go through have shown to have significant correlations with heart rate and RPE. However, all three assess different aspects of our training and we should take each into account. Potentially, looking at ratios of these variables and by starting to assess the intensities of the accelerometry data from our athletes (as it allows us to see the intensities in all planes of movement rather then GPS/camera derived accelerations and decelerations), we can then look to improve our interpretations of the athletes movement characteristics.

Inertial movement sensors now also detect jumps, caution must be taken when analyzing this data. Yes, using the flight time and initial acceleration we can calculate jump height, using gyroscopes to ensure the orientation of the unit is in the correct position for a jump to actually take place, but we need to take care. Comparing accelerometers and force platforms, we have seen the accelerometer (between the athletes scapulae) overestimates the jump height in comparison to the force platform, however the data is reproducible, a likely result of the accelerometers technology and its positioning on the body, in comparison to when we analyze jump height from ground reaction forces. However, it provides potential scope to assess neuromuscular fatigue in the field, while assessing more ecologically valid jumps during actual match play. It should be considered, that future research may want to investigate the use of accelerometry data within different planes of movement, quantifying accelerometry intensity thresholds and potentially assess accelerometers at different body locations. For example, if we have a player returning from injury with a lower limb injury, would wearing the unit between the shoulder blades, as what is common practice in elite sports, be the most appropriate location to assess if the injury has fully recovered and is efficiently ‘loading’ itself again? One argument about the use of accelerometers is commonly the noise coming from them due to their high sampling frequencies which puts into question their use. However, if we correctly apply the accelerometers, ensure correct practice takes place to reduce the noise (using the correct filtering system, reducing artefact noise) then the accelerometers may provide us with a new scope to investigate to aid our athletes performance demands and aid injury prevention.

Hopefully this brief review illustrates that we are just beginning to investigate the use of accelerometry within team sports. More detail is required as we look to investigate the relationships of these metrics to other commonly used training load variables. Three papers are currently available which assess the relationship of accelerometry data to RPE (Gomez-pirez et al., 2011; Scott et al., 2012) and heart rate data (Montgomery et al., 2010). Other sections on this website cover the issues with RPE and some heart rate methods. So are comparisons with these methods valid? If in your opinion they are then why not just use those methods, what novelty and advantage does accelerometry bring? Should we be looking to use accelerometry as a surrogate measure for internal load at all, given that similar utility movements between players, which may give similar accelerometry derived loading scores may not fully take into consideration the individual characteristics of the athlete (see the training process). I look forward to posting more articles in the coming months on more specific uses of accelerometry data which may help answer these questions.