Sport Informatics and Analytics/Performance Monitoring/The Quantified Self

Introduction

This topic explores two issues raised in Performance Monitoring, Theme 3 of this course.

1. The processes of quantifying performance.
2. Ethical issues raised by the surveillance of athlete performance and the quantification of self.

Quantifying performance

Technological innovation has made it possible to monitor sport performance (other-tracking) in unobtrusive ways. Jonas Lutz and his colleagues (2019) showcase "an overview of the environments in which the wearables are employed". The paper "elaborates their use in individual as well as team-related performance analyses with a special focus on reliability and validity, challenges, and future directions".

The growth in personal fitness and well-being devices has made it possible also to track one's own performance (self-tracking). Shona Halson provides a comprehensive overview of a range of approaches that combine other- and self-tracking measures of training load. Her overview of monitoring external and internal training loads contributes to the discussion of the use made in quantifying performance of objective and subjective measures of well-being . Shaun McLaren and his colleagues (2017) have contributed to this discussion with their meta-analysis of the relationships between internal and external measures of training load and intensity in team sports. Toby Edwards and his colleagues provide a meta-review of monitoring and managing fatigue in basketball. Dhruv Seshadri and his colleagues (2019) reported on wearable sensors for monitoring the physiological and biochemical profile of athletes.

One of the outcomes of a 2016 conference on monitoring athlete training loads was a 2017 consensus statement that provided "a summary of the key research and practical themes presented at the conference" and shared "directions for future developments in training-load monitoring". Arno Knobbe and his colleagues provide an example of such monitoring with their discussion of training schedules for speed skating. Carl Foster and his colleagues evaluate the "session rating of perceived exertion (RPE) method to quantitate training". Aaron Coutts and his colleagues provide an example of the use of RPE during football specific aerobic exercises. Jace Delaney and his colleagues (2018) provided examples of wellness data collected from female collegiate association football players. Marti Casals and Rasmus Nielsen (2019) have discussed improvements to the use of statistics in sports injury. They have looked at the ways in which multidisciplinary teams use the research evidence.

Shona Halson defines 'external load' as "the work completed by the athlete, measured independently of his or her internal characteristics". Measures of external load include: power output, speed and acceleration; time-motion analysis; and neuromuscular function.

She characterises 'internal load' as "the relative physiological and psychological stress imposed". Measures of internal load include: ratings of perceived exertion; heart rate; training impulse; lactate concentrations; biochemical, hormonal and immunolgical assessments; questionnaires and diaries; psychomotor speed; and sleep patterns. Wei Gao, George Brooks and David Klonoff explore how this monitoring of internal load occurs when wearable sensors are used to continuously detect target analytes in skin interstitial fluid (ISF), tears, saliva, and sweat. Brendan Lazarus (2017) and his colleges discussed a novel measure (weekly load derived from a weighted combination of Global Positioning System data and perceived wellness over a 24-week season) of training load in Australian rules football.

One extension of this work is the development of processes to report on and identify injury risk. Mladen Jovanovic discusses how these processes can become more robust and shares examples (2018a, 2018b) of how data can be prepared to estimate injury prediction. Michael Drew and Caroline Finch have added to this discussion with their consideration of the relationship between training load, injury, illness and soreness (see also Liam Toohey and colleagues' (2019) discussion of injury categorisation). Alessio Rossi and his colleagues provide a case study of injury prediction in association football that uses GPS data and machine learning. David Carey and his colleagues consider whether training load monitoring data could be used to predict injuries in elite Australian football players but note "injury prediction models built using training load data from a single club showed poor ability to predict injuries when tested on previously unseen data". Rasmus Nielsen and his colleagues discuss 'training load', 'structure-specific load' and 'load capacity' in their discussion of training and injury risk. Machar Reid and his colleagues (2018) revisited discussions about the reporting of tennis injuries in the light of "the streams of data now available" about players' workload. Alden Gonzalez and Gary Wilkerson, Ashish Gupta and Marisa Colston (2018) provide examples of the process of player monitoring and surveillance in American football. David Carey and his colleagues (2018) investigated the modelling of training loads and injury in Australian rules football and association football. Michael Bergeron and his colleagues (2019) reported on a supervised machine learning-based approach in modeling that estimated symptom resolve time in high school athletes who incurred a concussion during sport activity. David Carey and his colleagues (2018) examined the use of predictive modelling in Australian rules football and noted "injury prediction models built using training load data from a single club showed poor ability to predict injuries when tested on previously unseen data, suggesting limited application as a daily decision tool for practitioners".

Sam Robertson (2019) argued "Injury prediction is a waste of time". He noted fourteen problems in the area. One of his observations was:

Injury prediction models just aren’t implementable in practice. Data isn’t analysed, nor are models run - in near real-time. Models run pre- or post- session or competition cannot account for what happens during sessions – which is of course when the actual injuries occur.

Evangelia Christodoulou and her colleagues (2019) have drawn attention to the methods used to analyse data in clinical prediction models. Evangelia and her colleagues found no evidence of superior performance of machine learning over linear regression for clinical prediction modeling. They proposed that "improvements in methodology and reporting are needed for studies that compare modeling algorithms".

In 2017, Microsoft announced the development of a Sports Performance Platform that includes a machine-learning approach "to uncover insights around athlete readiness and injury prevention". In October 2017, it was reported that the Major League football team Houston Dynamo had partnered with Kitman Labs "to quantify and manage athlete risk, ensuring the best overall team performance across a long and often game-congested season". The report noted "since deploying Kitman Labs, the club has seen 63 percent fewer injuries and 75 percent fewer days lost to injury. The club also experienced 88 percent reduction in strains and sprains". General Electric worked with the International Olympic Committee at the Rio Games in 2016 to provide a cloud-based electronic medical record system. General Electric was contracted to provide an Athlete Management Solution for the Olympic Winter Games in PyeongChang in 2018 and the Olympic Games in Tokyo in 2020.

Anne Simpson and her colleagues report the use of "an image-based food record and social-media functionality to provide in-application personalised feedback to individuals or groups, peer-support, and a platform to deliver nutrition education material". Athletes used a proprietary smartphone software tool to log images of their meals three days per week over a six-week period. Martin O'Reilly and his colleagues have outlined how an app for biofeedback can assist physiotherapists and personal trainers in exercise monitoring.

The comprehensive monitoring of training performance enables coaches to modulate training. Daniel Plews and Paul Larsen provide an example of this modulation in their discussion of an Olympic rowing crew.

Daniel Alvira and his colleagues (2018) provided an example of the quantification of internal load of association football referees. (See also, Daniel and Castillo and colleagues' (2016) discussion of internal and external load indicators for association football match officials).

In 2018, the Physical Activity Guidelines Advisory Committee summarised the scientific evidence on physical activity and health. I was used and to develop physical activity guidelines for Americans. The report included a section on physical activity behaviours: steps, bouts and high intensity training.

Jeff Griesemer (2019) provided details of using Fitbit data and R to create timeline charts. I-Min Lee and her colleagues (2019) reported on the association of Step Volume and intensity in a cohort study of 16,741 women with a mean age of 72 years, over a 7 day period.

Meta reviews

Using integrated technology in team sports

Carla Dellasera, Yong Gao and Lynda Ransdell have provided a narrative qualitative review of "IT's emerging impact in sport settings". Their paper reviews 39 publications and has 83 references. As part of your investigation of the quantified self, can you read this paper and reflect on the benefits of accessing an authoritative meta review? Are there any issues in learning about research activity without accessing the primary source? You might extend this reflection by reading Shaun McLaren and his colleagues' (2017) meta-analysis of the relationships between internal and external measures of training load and intensity in team sports.

The quantified self

The Wikipedia page on the Quantified Self provides a detailed overview of the emergence and development of the Quantified Self movement.

It includes references to Gary Wolf and his view of the data-driven life. His overview of the emergence of technologies to make self-tracking more accessible includes this observation:

In the past, the methods of quantitative assessment were laborious and arcane. You had to take measurements manually and record them in a log; you had to enter data into spreadsheets and perform operations using unfriendly software; you had to build graphs to tease understanding out of the numbers. Now much of the data-gathering can be automated, and the record-keeping and analysis can be delegated to a host of simple Web apps. The makes it possible to know oneself in a new way.

Kevin Kelly suggested in his preliminary discussion of the Quantified Self:

Unless something can be measured, it cannot be improved. So we are on a quest to collect as many personal tools that will assist us in quantifiable measurement of ourselves. We welcome tools that help us see and understand bodies and minds so that we can figure out what humans are here for.

Melanie Swan defines the quantified self as:

Any individual engaged in the self-tracking of any kind of biological, physical, behavioral, or environmental information.

Some contributors to the discussions about the Quantified Self have characterised self-tracking as Personal Informatics. and the 'laboratory of the self'. Other commentators caution about the reliability and accuracy of data captured by self tracking devices and how these data might be visualised "in useful and meaningful ways". In 2017, Btihaj Ajana and her colleagues shared a film of the Quantified Life that focussed on the self-tracking practices of Thomas Christiansen. Eric Finkelstein and his colleagues (2016) and John Jakicic and his colleagues (2016) considered the impact of wearable technologies on behavioural change. Farhad Manjoo (2019) reported on his use of trackers as part of the New York Privacy Project.

Deborah Lupton has explored the interface of critical social research and human-computer interaction (HCI) in personal informatics. (See also her more recent work (2019) that takes "a feminist new materialist perspective, focusing on relational dimensions, affective forces and agential capacities".) Her 2019 book discusses how people make sense of and use their personal data, and what they know about others who use this information.

A body of literature has now been established of research that has sought to investigate the social, cultural and political dimensions of self-tracking, nearly all of which has come out in the last few years. This literature complements an established literature in human-computer interaction research (HCI), first into lifelogging and then into self-tracking (or personal informatics/analytics, as HCI researchers often call it).

Deborah Lupton explored contemporary self-tracking cultures in her book published in 2016 and developed the concept of 'lively data' "to denote the manifold ways in which personal digital data ... are vital" (see also, the discussion of 'mundane data'). In 2017, Deborah and her colleagues discussed self-tracking, health and medicine and agential capacities. In 2018 Deborah discussed lively data in the context of the intersections of health and fitness self-tracking and social media.

Ben Williamson has used the term 'algorithmic skin' to discuss "the ways that health-tracking produces a body encased in an ‘algorithmic skin’, connected to a wider ‘networked cognitive system’". The emergence of implantable, non-medical monitoring devices moves this algorithmic skin to a subcutaneous level and raises issues about the use and misuse of such devices. Haley Weiss (2018) provided an extended discussion of microchip implants.

Deborah Lupton and Ben Williamson are examples of a reflexive approach to "data-led and algorithmically mediated understandings of the body". Antii Poikola, Kai Kuikkaniemi & Harri Honko have extended this discussion to explore " the right of individuals to access the data collected about them". Marjolein Lanzing has added to the debate by considering how self-trackers negotiate the 'extended transparency' of their practices. She points to "tension between the idea that one should disclose personal information in order to gain more self-control and the informational privacy one needs to live an autonomous life". Kate Crawford, Jessa Lingel and Tero Karppi note "wearable self-tracking technologies reflect the simultaneous commodification and knowledge-making that occurs between data and bodies". They explore "agency, practices of the body, and the use of wearable data by courtrooms and data science to enforce particular kinds of social and individual discipline". Junaid Mubeen explores another aspect of our relationship with quantification and the autonomy we have as app users. Andrew Jackson (2018) has discussed privacy issues related to the use of wearable technologies in respect of Australia’s federal privacy legislation.

Rob Horning considers issues around 'algorithmic identity' and the challenges we face in being ourselves and managing the "unremitting pressure" in order "to make the most of ourselves and our social connections and put it all on display to maintain our social viability". Our connections with others in urban contexts raises questions about how our data might inform urban design. Minna Ruckenstein and Natasha Dow Schüll (2017) with their consideration of datafication, Minna Ruckenstein and Mika Pantzar's (2017) discussion of a dataistic paradigm, and Btihaj Ajana's (2017) exploration of "the emerging tension between philanthropic discourses of data sharing and issues of privacy" have extended this discussion. (See also Btihaj Ajana's (2018) analysis of metric culture and Vaclav Janecek's (2018) discussion of the ownership of data in the Internet of Things.)

There is a growing industry around algorithmic identity that seeks to personalise "fitness coaching at scale". These services include: genomics fitness recommendations; 3D body scanning technologies to track personal fitness journeys; and predictive analytics and artificial intelligence enabled coaching. This trend raises further questions about access to personal genomic data that "reveals intimate information about not only you, but also the people to whom you are related". In 2015, a consensus statement on direct-to-consumer genetic testing for predicting sport performance stated:

The general consensus among sport and exercise genetics researchers is that genetic tests have no role to play in talent identification or the individualised prescription of training to maximise performance.

Debates about genomic data are discussed in detail in the Australian Council of Learned Academies report (2018) The Future of Precision Medicine in Australia. They are the subject of an editorial in the British Medical Journal (2017) also.

Dhruv Seshadri and his colleagues (2019) reported on wearable sensors for monitoring the physiological and biochemical profile of the athlete. They noted "The emergence of flexible and stretchable electronics coupled with the ability to quantify biochemical analytes and physiological parameters have enabled the detection of key markers indicative of performance and stress".

Gabija Didziokaite Paula Saukko and Christian Greiffenhagen (2018) interviewed ‘everyday calorie trackers’ and highlighted "the mundane side of self-tracking, where people pursuing everyday, limited goals engage in basic self-tracking and achieve temporary changes".

Erman Misirlisoy (2019) discussed polysomnography (a direct scientific assessment of sleep) and sleep trackers. He noted "it’s also hard to gauge a particular model’s reliability". Azizi Seixas and colleagues (2019) examined sleep duration and quality from 2,194,897 users of a popular sleep tracking application over a four-year period (2015 to 2018). They concluded "our findings from big data are consistent with previously reported estimates of sleep duration and quality. Sleep duration varied by age, sex, day of the week, and season".

Resources

Links

Some of the resources to inform our discussions about the quantified self can be found in this mind map. Recommended and suggested reading on this topic can be found in this wiki

.

Global positioning system

Introduction

James Malone and his colleagues (2017) note "athlete-tracking devices that include global positioning system and microelectrical mechanical system components are now commonplace in sport research and practice". These devices are fundamental to the quantification of other and self performance and, as Robert Aughey points out, owe their precision to the development of nuclear magnetic resonance method and the creation of atomic clocks.

The Global Positioning System (GPS) we use today is the outcome of research and development initiatives by the United States military in the 1960s and 1970s. The system was made available for public use in the middle of the 1980s following a decision made by the United States Government in 1983. It is a multi-use, space-based radio navigation system owned by the United States Government and provides users with positioning, navigation, and timing services. It is comprised of three segments: the space segment, the control segment, and the user segment. The United States Air Force develops, maintains, and operates the space and control segments.

GPS in sport research and practice

James Malone and his colleagues (2017) observe that "the use of GPS in sport allows practitioners to evaluate athletic training programs, and researchers to better investigate applied research questions". Commercially available GPS devices have integrated micro electrical mechanical systems, such as triaxial accelerometers, magnetometers, and gyroscopes, into their platforms. James and his colleagues note that it is important that researchers and practitioners are aware of the how these data are derived from these systems. They add:

More specifically, it is important to understand how these data are generated, the factors that affect measurement validity and reliability, the impact of changes in hardware/software and how data should be reported.

One of the earliest papers in the literature on GPS and physical acitivity was written in 1997 by Yves Schutz and A. Chambaz . They asked Could a satellite-based navigation system (GPS) be used to assess the physical activity of individuals on earth?. Their study used one subject and investigated data from a GPS unit with chronograph measurements. They compared nineteen measurements for walking speed and twenty-two measurements of running speed. Their paper identifies advantages and disadvantages of using a GPS system.

Advantages:

• Portable (light and small size)
• Non-invasive, non-obtrusive, free-living measurements
• Continuous measurement available on miniature screen
• Free access to GPS satellites
• Reasonable costs of a GPS receiver
• Data storage and retrieval
• Independently validate other measurements

Disadvantages:

• Only activities involving outside displacements of the body, such as walking and running can be assessed
• Failure to measure displacements when the access to the sky is obstructed by tall buildings or terrain
• Static activities cannot be measured
• Depends upon the continuous access to at least three satellites simultaneously (for 2D assessment).

In a subsequent paper (2000), Yves Schutz and his colleagues discussed the use of GPS in the study of the biomechanics of human motion. Issues with the accuracy of GPS were addressed in another early paper (2001) written by Peter Larsson and Karin Hendriksson-Larsen. They considered the use of differential GPS (dGPS) "to investigate whether it would be possible to relate dGPS data with physiological variables in a field test and thus achieve a more controlled field test and hence improve the value of sport-specific testing". Peter and Karin concluded "dGPS was shown to be a method that could give detailed information about a subject's speed and position" and "physiological variables could be related to dGPS data". Peter wrote a second paper about the use of dGPS and sport-specific testing in 2003.

Aaron Coutts (2017) discussed the theoretical basis and practical application of athlete tracking technologies and with his colleagues (2017) considered the black box characteristics of GPS devices..

Simon Legg (2017) discussed the use of GPS in Australian rules football in conversation with a high performance manager.

For an example of the visualisation of GPS data in cycling and adventure sports, see Data61's use of Doarama a 3D visualisation engine for geolocated activities. Andrew Novak has explored the use of GPS visualisation in invasive team games.

Darcy Brown and his colleagues (2017) researched "the validity of a global positioning system (GPS) tracking system to estimate energy expenditure (EE) during exercise and field-sport locomotor movements". They concluded "a GPS tracking system using the metabolic power model of EE does not accurately estimate EE in field-sport movements or over an exercise session consisting of mixed locomotor activities interspersed with recovery periods; however, is it able to provide a reasonably accurate estimation of EE during continuous jogging and running".

Jonathan Shepherd and his colleagues (2018) provided an overview of inertial sensor propulsion measurement in wheelchair court sports.

Varuna De Silva and his colleagues (2018) shared their analysis of 53 association football players in a football academy based on data collected over four years that provided 20, 913 data points. Their aim was "to analyse player tracking data to understand activity level differences between training and match sessions, with respect to different playing positions". Adrian Gray and his colleagues (2018) discussed their use of GPS data to improve the assessment of mechanical load in contact team sports.

In January 2019, the Football Technology Innovation department at FIFA reported a tracking systems validation study conducted in November 2018.

The Football Technology Innovation department at FIFA conducted a research study with Victoria University at the Mini Estadi, in Barcelona supported by the Barça Innovation Hub to explore the validity of 16 different Electronic Performance Tracking Systems (EPTS). The systems, made up of Global Positioning Systems (GPS), Local Positioning Systems (LPS) and Optical Systems were tasked with tracking 10 players through specific football movements and small sided games. Positional and velocity data were collected by all providers which is now being compared with the processed VICON tracking data to assess the levels of agreement between the two.

The aim of the project was "to understand the accuracy and create a more transparent overview of available systems and their accuracies as the global demand for tracking systems increases, with the final goal to create a professional standard" (our emphasis). There is a video of the procedures used for this study.

Martin Buchheit and Ben Simpson (2017) noted "the decision to use any tracking technology or new variable should always be considered with a cost/benefit approach (ie, cost, ease of use, portability, manpower/ability to affect the training program".

Reflecting on the use of GPS

To get a feel for the literature on the use of GPS, we suggest you have a look at the two early papers on GPS and two on dGPS in the context of James Malone and his colleagues' (2017) meta-review of the literature, particularly with regard to the validity and reliability of the data collected by these systems. You might like to refer to Robert Aughey (2006) (and his 2018 work with FIFA on electronic performance and tracking systems), Aaron Coutts and Rob Duffield (2010), Cloe Cummins and her colleagues (2013), Daniel Link and Martin Lames (2018) for their systematic reviews of GPS and microtechnology sensors, Daniel Nicolella and colleagues' (2018) discussion of the validity and reliability of an accelerometer-based player tracking device, and Heidi Thornton and her colleagues' (2018) consideration of inter-unit reliability and effect of data processing methods of global positioning systems. For a discussion of the use of heart rate data variability in measuring devices see Marco Altini (2018) (for further discussion of hear rate variability see Marco Altini, 2017).

We recommend that you read Alec Buttfield's (2018) account of his engagement with GPS systems in his role as a sport scientist at an institute of sport. In it he discusses the early adoption of GPS systems and how to use the data produced by the systems. See too, the toolkit Alec developed to analyse athlete tracking data.

You might also find a generic paper on biologging of interest. Rocio Joo and her colleagues (2019) reviewed 57 R packages used to track movement and identified 12 packages with good or excellent documentation.

Relative use of speed zones

After reading about the use of GPS and dGPS, we think you might find discussions about relative speed zones of interest. Researchers investigating performance in Australian rules football, rugby league, rugby union and women's rugby union sevens have considered the use of the individualisation of velocity thresholds and the prescription of training load to optimise performance in game contexts.

Question What do you think are the merits of using relative rather than absolute measures of sprint performance?

Case studies

Case study 1: association football

Does training affect match performance?

We hope this case study might help us explore how the use of tracking devices can lead to informed discussions about the relationship between training environments and performance in competitive games.

Javier Fernández and his colleagues reported the use of an electronic performance and tracking system (including GPS and microsensor technology) at FC Barcelona. Their paper shares data from the 2015-2016 season for the FC Barcelona B team. In the paper, Javier and his colleagues discuss:

• Data collection
• Data processing
• Data exploration

When you read their paper you might consider the domain specific knowledge used to construct the process of study and how this approach might be used with other sports.

Note: If you would like to learn more about Javier Fernández's work, he completed his Masters thesis in 2016 at the Universistat Politecnica De Catalunya. The title of the thesis is From training to match performance: an exploratory and predictive analysis on F.C. Barcelona GPS data.

.

Case study 2: outdoor sports

MAProgress

We hope this case study might help clarify the discussion of quantification of performance.

A New Zealand company, MAProgress, provides a real-time GPS service for outdoor sporting events. Their website observes:

Our live tracking system helps event directors run high-profile events by utilising the latest in GPS satellite and digital technology. Make your event more interactive, get more exposure for sponsors and create an experience that participants can share with friends and family from afar. It’s about safety, logistics and communication.

This video illustrates the service provided to a coast to coast race in New Zealand. The company uses these resources to deliver these features. The MAProgress platform was used for the Indian Pacific Wheel Race 2017. An accident towards the end of the race raised some fundamental issues about the interplay between personal data and public interest.

For a discussion of MAProgress and other GPS applications in adventure events see Stephen Schroeder (2017).

Case study 3: personalised nutrition

Predicting glycemic responses

David Zeevi and his colleagues devised a machine learning algorithm to predict personal postprandial gylcemic response to food. Their paper shares their methodological approach to using personal profiles in computational analysis. In doing so, they raise some profound issues about how quantification (in their case through a continuous glucose monitor) of personal responses can lead to changes in individual behaviour through disciplined use of machine learning. We recommend that you explore this paper in detail to consider how you might adapt your own approach to the quantification of performance.

.

Case study 4: collecting and visualising personal data

Dear-data and dear-data-two

Giorgia Lupi and Stafanie Posavec collected and measured different aspects of their lives for a whole year and shared their data with each other as data visualisations on postcards. "Eventually, the postcard arrived at the other person’s address with all the scuff marks of its journey over the ocean: a type of “slow data” transmission."

Giorgia and Steanie noted in their discussion of the project:

We prefer to approach data in a slower, more analogue way. We’ve always conceived Dear Data as a “personal documentary” rather than a quantified-self project which is a subtle – but important – distinction. Instead of using data just to become more efficient, we argue we can use data to become more humane and to connect with ourselves and others at a deeper level.

Their correspondence has been collected in a journal (2018).

Jeffrey Shaffer and Andy Kriebel undertook a dear-data-two project inspired by Giorgia and Stafanie's work. Jeffrey and Andy said of their project:

We are both data visualization practitioners, teachers and regular speakers on the topic and we are big proponents of data visualization best practice. Giorgia and Stefanie are amazing artists, which is evident in their work. We are technology guys, using tools like Tableau and R to create data visualizations. We thought it would be fun and interesting to see the differences in the approach and design of the data and contribute to this amazing project.

We suggest you spend some time exploring both of these projects. They are excellent examples of the connections made in this course across our four themes: introductions, pattern reconition, performance monitoring, and audiences and messages.

Case study 5: visualising a daily commute

From home to the work place

Matt Grobis was inspired by the Dear Data books and decided, with friends, to collect data over a three-week period about transportation and travel.

I decided to quantify my walks to work. I used the GPS Logger for Android app to get 1 Hz time series data for 44 commutes. Data collection began the moment I left my apartment building or work, and ended when I touched the door handle of my destination.

Matt has shared his R code for this project on Github. If you are interested in using a logger for your daily activities, we recommend Matt's project to you. His account includes the visualisation of data and demonstrates how to anonymise the data on a map.

Case study 6: wrist-worn accelerometers

Field evaluation

Toby Pavey and his colleagues (2017) discussed data collected with a wrist-worn accelerometer (a GENEActiv monitor). The aims of their study were to: determine if models trained on laboratory data perform well under a 24 h free-living conditions; and train and test a random forest activity classifier for wrist accelerometer data. Their paper resonates with a number of themes in this course including performance monitoring and pattern recognition. Toby and his colleagues used the CRAN package randomForest. We recommend that you read their paper and reflect on their conclusion in your own use of accelerometer data:

The random forest classifier for wrist accelerometer data yielded accurate group-level predictions under controlled conditions, but was less accurate at identifying stepping verse non-stepping behaviour in free living conditions.

Do you think adding more accelerometers, for example to the hip, might help?

Case study 7: sleep patterns

Fitbit data

in January 2018, Fitbit released data from six billion nights of sleep. Fitbit refined its sleep data collection process in March 2017. At that point, the company started to measure sleep stages to capture information about light sleep, deep sleep and rapid eye movement. It sought to gain sleep insights "to discover trends about what may be affecting your sleep and then offer up personalized guidance on how to improve it".

We suggest you look at the two discussions of the data by David Pogue and Alex Gray to get a feel for the data collected. You might also like to read the American Academy of Sleep Medicine's consensus statement on sleep duration recommendation and Thomas Penzel and his colleagues' (2003) discussion of heart rate and sleep stages.

Chris Davidson (2018) provided an example of the process of collecting sleep data and integrating the insights into personal lifestyle choices.

Case study 8: respiratory frequency

Monitoring training

Andrea Nicolo, Carlo Massaroni and Louis Passfield (2017) noted that respiratory frequency (fR) was emerging as a valuable measurement for training monitoring. Their paper provides a comprehensive discussion of respiratory frenquency. They conclude "respiratory frequency represents a good example of how wearable sensor development should follow athlete's needs and be informed by scientific findings".

We recommend that you look at this paper and reflect on the ways in which you collect data to monitor performance. We think it might prompt you to think about the decisions you make about what data to collect.

Case study 9: cricket bowling

Using sensors to monitor bowling actions

Andrew Wixted and his colleagues have explored the use of wearable sensors to detect bowling actions. Their papers give an insight into the processes used to monitor movement actions in real world settings.

Transparency

Gina Neff and Dawn Nafus (2016) note:

Self-tracking is a human activity, one far more interesting than the gadgets that have made it easier and more widespread. Self-tracking does not necessarily require technology more complex than pen and paper. However, much self-tracking is now digital, whether done via wearable computers ... These technogies intersect with the ways that people have self-tracked for centuries like keeping diaries or logs. The growth of these digital traces raises new questions about this old practice.

Marjolein Lanzing (2016) addresses one of these questions. She proposes:

while self-tracking may sometimes prove to be an adequate method to shed light on particular aspects of oneself and can be used to strengthen one’s autonomy, self-tracking technologies often cancel out these benefits by exposing too much about oneself to an unspecified audience, thus undermining the informational privacy boundaries necessary for living an autonomous life.

The availability of self-tracking devices makes it possible to make our own activities transparent to ourselves and to others. Marjolein Lanzing observes:

Contrary to a written diary, the terms and measures we employ to self-monitor are not selected by the user, but part of the design of the device. A self-tracker cannot control or be sure that third parties will not access her data. Ignoring the change of the cultural practice along with new technological potentialities of self-tracking devices contributes to misconceptions about the way information is collected, shared and stored.

As you reflect on the technological opportunities for tracking (self and other), you might consider how the transparency of quantified performance impacts on your own practice and the decisions you make about the privacy of personal data.

Reflections

Technocratic engineers?

This course looks explicitly at ethical issues related to monitoring and surveillance. Andrew Manley and Shaun Williams observe

Coaches often voice a humanistic approach to their practice; however, with an increased reliance on surveillance and data capture, it may be that we are visibly witnessing the caricature of the elite coach morphing into the modern-day technocrat engineer. And while it can appear to drive performance, at such an unrelenting pace it may not be sustainable in the long run".

Do you share their concerns? Does an article about players' right to data by Ian Cameron (2018) help you in your deliberations?

Cost-benefit of monitoring?

Chris Carling and his colleagues (2018) have invited us to contemplate the relevance of "a comprehensive body of research" that has investigated post-match acute and residual fatigue responses in association football. We recommend you read their paper in order to reflect on the process of performance monitoring. What are the implication for practice of their suggestion:

Fatigue monitoring requires a more practical approach using data derived in training sessions andthe development of tools to enable the simultaneous, instantaneous and non-invasive capture of multiple sources of information during and following play.

Global thresholds and individual capacities?

Will Abbott and his colleagues(2018) shared data from 31 professional association football players in an English Premier League academy. We recommend that you review this paper to learn how the authors sought to reconcile global thresholds for acceleration with the individual differences in the performance of the 31 academy players monitored in ecologically valid contexts.

In 2019, the University of Brighton advertised a PhD scholarship to investigate "tracking and monitoring player health, well-being and training load in elite football players". Whilst not directly related to the acceleration of players, this scholarship aimed to "investigate athlete wellness, encapsulating physical, physiological, and psychological markers, to determine athlete readiness to train, athlete readiness to play, and potential injury risk". The multidisciplinary approach would have to address Will Abbott and colleagues' discussion of acceleration albeit with a different cohort of players possibly within the same football club with the opportunity to develop long-term profiles of players' performances.