Influences of GPS & Accelerometry on Practice and the Training Process

First of all apologies for not really writing an article for a while, been busy developing our athlete monitoring software (enough said about that for more info).

I’ve been inspired to write this article mainly by a conference I attended organised by Perform Better and Steve Barrett (who has written for this website before on the use of GPS and accelerometers) but also a number of discussions I’ve had with sports scientists around the UK about their use of athlete tracking technology mainly in soccer and rugby. Credit must be given to Perform Better, Catapult and all the presenters as keeping an open forum such as this may lead to criticism and some uncomfortable questions. But I think and hope I speak for all of them that they do this to share ideas, pick brains and to develop our understanding of the area as a whole. Many of the presenters and the organisers alluded to this as one of  the purpose of the conference.  This article is about facilitating that process and why this website exists. At the conference a number of presentations were delivered through the day by practitioners from elite football (soccer) clubs and a rugby club. I may not mention where all of them were from as I have not have had express permission to mention the clubs or people although I’m sure most of them wouldn’t mind. For those of you not familiar with the website and my background I encourage you to read about the training process. I firmly believe in that model proposed by Impellizerri but think it’s probably useful to add some more information to that model for practitioners. As I learn more I hope to develop and refine that model so that we can start to better understand the training process in relation to some of the newer technology that is becoming available. This newer technology will be the focus of this article and I hope to address and open for discussion three main areas.

  • Applications of athlete tracking technology.
  • Validity of Measurements.
  • Where does it fit into the training process?

Applications of athlete tracking technology.

Steve Barrett very kindly covered some of the applications of GPS and accelerometers in his article a few months back on this website. You can measure distance which is useful but this may not accurately represent movement in all three planes where tri-axial accelerometers pick up additional information. The GPS traces can also give an indication of accelerations and decelerations. This is all information we haven’t always been able to get before therefore now more that ever we can assess the movement demands of specific sports to a greater degree than ever. Here I will differentiate movement demands to metabolic demands. In future after some more consultation with those doing research in the area of metabolic calculations from GPS data will post another article specifically on this. But for the purposes of this article movement demands are the demands expressed by measurement of movement through either GPS and /or accelerometers. So according to the training process these measurements will determine the internal load in relation to the individual’s characteristics. Ultimately it is the internal training load that will determine the training outcome, so is the data useful for anything?

Well there was some effective application of some of this data presented at this conference and some of this I know from coaches in rugby league. There was one approach employed where the percentage of acceleration in each plane was mapped for each training session. This was fairly consistent until the player concerned picked up a back injury. At this point the distribution of acceleration in each plane changes markedly and as the player returned to full fitness the distribution gradually returned to what was normal for him when fully fit. Changes in gait and therefore lateral & vertical accelerometer readings in the vertical and lateral planes may prove to be useful in detecting injuries or for rehabilitation purposes. This was also a great example of assessing a player to his own normative values, an approach that must be adopted more widely if we are to truly understand the individual.

Rob Heyworth from Blackburn Rovers FC showed some interesting data on how manipulation of playing area affected accelerations and decelerations. One of the most important findings with his data for me was the individual variation in the number of these action in each drill. Specific running drills still have place in training for Matt Reeves at Leicester City FC. He’s used his own developed drills to make sure players reach the external intensity & frequency required rather than leave it to chance in a small sided game. Matt argues that heart rate responses can mask how hard the players are working when at high speed so GPS and accelerometers give him the insight HR doesn’t. This is a view supported by many and one I can fully understand especially in very acute bouts of exercise. On the other hand when the speed is zero, heart rate will still be elevated. My belief is that at very high speeds there may be underestimation of intensity instantaneously with heart rate, but the internal load accrued from the elevated HR between bouts somewhat reduces this underestimation and the aerobic contribution to repeated sprint exercise has also been shown previously. The anaerobic component is difficult to factor in and one of the challenges faced although time spent at sprint intensity in actual match play is only 2%. On the other hand could it and should it be treated as a different entity and analysed as such.

Regardless of this issue it is another method of effective application of GPS data, where it allows movement demands and patterns to be examined and trained.  Newer developments in these devices with the integration of gyroscopes and magnetometers with GPS and accelerometers may give us more insight into movement demands in future. It will definitely allow insight into the direction a player is facing when they are moving and this may have implications for injury prevention, mapping turning patterns, subsequent training and injury prevention. So the technology has somewhat enhanced our understanding of movement demands. I am sure there are many other uses when it comes to tracking athlete movement that influences training regimens and if you would like to share yours please post on the comments section below.

One of the main ways a number of clubs presenting at the conference monitored their loading is as a percentage of match demands for the various components these technologies can measure. I am still unsure of the rationale for this and no one has been able to provide me a good reason as to why this is done. Is there a scientific underpinning to this or is it just because we can measure match demands so we will use it as a yard stick. For me for it to be viable there must be a dose-response relationship where load equivalent to a certain number of games produces a proportional response and I have not seen that evidence yet.  Add to that the variances between positions, within position, through systems of play, tactics etc, the match demands become blurred. Different approaches have been used to come up with the “match demand” from using friendly match data to reserve team data to create positional demands. A further question that arises if training is designed to mimic match play, does this provide the overload required to adapt and meet match demands comfortably? An approach presented at the conference by a rugby club was the idea of a “worse possible scenario”, where players are pushed beyond the measured demands of their sport. This would make sense from an overload and progression perspective. But if these practices are adopted solely on the use of external load measurements as an individual’s condition changes the internal load due to that external load will change and that will ultimately affect the training outcome not the percentage of game values or amount of external work per se. This is not to say those presenting at the conference were ignoring internal load.

So this leads me to question how these movement demands and measurement of movement translates into the actual physiological stress for each individual. Can it ever do this without assessment of the internal exercise stress whether it is mechanical (external) or cardiovascular (internal)?

Validity of Measurements.

 One of the measures of external load that I am yet to be convinced by is the measure of accelerometer derived load whatever the companies decide to term it. This measure is taken quite seriously by some and others I know who have used these system for over 5 years totally discard it. I conducted short impromptu twitter poll and I found that many practitioners and coaches have realised the limitations, are unsure of what the numbers mean or don’t care much about them.

In my opinion some attention has been drawn to the “term” load as this means we can periodize more efficiently and accurately. No doubt that companies have used this in their sales pitch in the early days of trying to sell these systems. I challenge the accuracy of these measures as a measure of load both in a mechanical sense and metabolic sense. It was great at this conference to get some evidence of something I’ve always theorized. Whenever I’ve been asked about the suitability of accelerometer derived load I give two scenarios the answers to which I think as it stands places accelorometry derived load in context. The 1st example is of squatting: An accelerometer measures the displacement in a given time in a given plane. What would it measure if I was to squat 100 kg at the same velocity as I squatted 50 kg? If the accelerometer measures what it is suppose to then it will give us the same value regardless of mass and the stress on our muscles. So this is an example of it not being able to discriminate within the individual on a mechanical basis mainly due to acceleration being measured independently of mass the product of which would give us a measurement of force.

I’ve also always theorized about accelorometry derived load on a wet pitch and dry pitch. I theorized on a softer pitch there would be more absorption and so the accelerometer would move less reducing the score. However for those of us that have ran or played on soft/wet pitches we find it much harder. Some evidence presented at this conference looked at the impact of ground hardness on accelerometer derived load. Softer ground increased internal load by elevated heart rate which would support the theory of it being harder but the accelerometer derived load didn’t change. Again accelerometry doesn’t discriminate in the way needed. To understand really why this happened all 3 planes of movement would have to be compared to see if any changes occurred or whether gait was maintained but required greater muscle recruitment hence greater oxygen demand and increased heart rate. 

Furthermore is every movement measured in a game situation or training a physiologically stimulating movement. Some of you may have data that can assess this! For example what is the load derived from a fall or a contact (usually defined by a spike on the graph) and what is the equivalent of that load from running? Does X amount of running = Y load from a fall? Are they both as physiologically stimulating? The answers to these questions will give us some insight and I look forward to your thoughts.  So that’s the mechanical side of it. In terms of the metabolic side of the argument the movement efficiency as a result of adaptation will always change within the individual. So as an athlete gets fitter and/or stronger the accelerometer load values would still be the same or the exact same exercise despite lower internal stress. There is some work being done on metabolic power derived from accelerometers and I will examine this in more detail in a future post.

Despite what I have written my aim is not to rubbish the use of accelerometers but just to raise awareness of the potential limitations, especially when used in isolation and examined in the wrong part of the training process as explained later in this article. Credit must go to these guys working hard with the data trying to make it effective and useful. Recently I also met with Richard Akenhead, sports scientist at Newcastle United FC. One of his doctoral studies examined the validity of accelerations from  GPS traces, you can find a poster of his research on Research Gate. In summary he found that accuracy in straight line accelerations at speeds above 4 m˙s-1 was compromised. This limitation must also be considered especially when we try to match or overload values from sport specific situations at the high end, the error involved may represent values that actually are not achievable.

There maybe other issues you are aware of or maybe you want to bring to attention of everyone or want to disagree with some of the statements I’ve made. Add comments below or I will happily publish a journal type rebuttal on the website.

Where do acceleromters & GPS fit in the training process?

Finally comes the training process, I have added some thoughts to the training process diagram that has helped me clarify where GPS and accelorometry may help to make an impact based on the research and practice I have come across. Again you may feel you want to add to this or change it. Although the applications are clearer its place in the training process remains. What I have seen and heard and somewhat disagree with is how all these load measures internal, external or subjective tell you different things and all must be looked at together. I think examining such data in that way may give us some insight but I have also seen that argument used to justify clumping external, internal load and subjective load together to provide a number. If (big if) all these numbers are valid, the summation of these must give a valid measure of “overall load” is the assumption I’ve seen made. According to the training process model the external load in relation to individual characteristics (genetics, fitness status) will give an internal load which will then determine the training response. It is not a simple addition or multiplication of each variable. If some of the practices I’ve seen were put into a model external load would be in the box where internal load is now. Some of this has stemmed from the relationships accelerometry derived load shows with supposedly valid measures of internal load. I would encourage you to examine the validity of those so called “valid measures of internal load”. This has been covered on the “training load” section of this website.

In summary the usefulness of accelerometry and GPS cannot be denied and it can impact directly on practice. The extent to which it impacts on practice accurately and efficiently is the area for debate. I feel sometimes it’s uses are overstated especially when internal load is partly or entirely replaced by external load in the training process model. Is equipment driving practice through overstated uses of “fanciful numbers” as it was put quite nicely put by one coach on twitter or scientific principles? Hopefully some of the studies I and others have planned over the next 12 months will give us answers that may help understand the direction we should be going in. A study of mine that was recently published integrating the internal and external load will be just one of many areas myself and fellow collaborators will be examining. If any of these area’s interest you feel free to get in touch.

The whole purpose of this website is to generate discussion and examine practice globally. So whether you agree or disagree share your thoughts in the comments section below. There are people out there with much more experience and insight on this area than me and I and all the followers of this website would love to hear from you.

Until next time…Thanks for taking the time to read this article.

If you want to post an article on this website feel free to contact me on


Dr Ibrahim Akubat