Content uploaded by Patrick Ward
Author content
All content in this area was uploaded by Patrick Ward on Feb 05, 2019
Content may be subject to copyright.
“Putting the ‘i’ Back in Team” by Ward P, Coutts AJ, Pruna R, McCall A
International Journal of Sports Physiology and Performance
© 2018 Human Kinetics, Inc.
Note. This article will be published in a forthcoming issue of the
International Journal of Sports Physiology and Performance. The
article appears here in its accepted, peer-reviewed form, as it was
provided by the submitting author. It has not been copyedited,
proofread, or formatted by the publisher.
Section: Invited Commentary
Article Title: Putting the ‘i’ Back in Team
Authors: Patrick Ward1, Aaron J. Coutts2, Ricard Pruna3, and Alan McCall4,5
Affiliations: 1Seattle Seahawks, Seattle, WA, USA. 2Faculty of Health, University of
Technology Sydney (UTS), Australia. 3FC Barcelona, Barcelona, Spain. 4Arsenal Football
Club, London, UK. 5Edinburgh Napier University, Edinburgh, UK.
Journal: International Journal of Sports Physiology and Performance
Acceptance Date: June 12, 2018
©2018 Human Kinetics, Inc.
DOI: https://doi.org/10.1123/ijspp.2018-0154
“Putting the ‘i’ Back in Team” by Ward P, Coutts AJ, Pruna R, McCall A
International Journal of Sports Physiology and Performance
© 2018 Human Kinetics, Inc.
Title: Putting the ‘i’ back in team
Authors: Patrick Ward1, Aaron J Coutts2, Ricard Pruna3, Alan McCall4,5
Institution:
1 Seattle Seahawks, Seattle, USA
2 University of Technology Sydney (UTS), Faculty of Health, , Australia
3 FC Barcelona, Barcelona
4 Arsenal Football Club, London, UK
5 Edinburgh Napier University, Edinburgh, UK
Corresponding author:
Alan McCall
Arsenal Football Club
Bell Ln, London Colney
Hertfordshire, AL2 1DR
Tel: +33 651748266 - Fax: +33 320887363
Email: amccall@arsenal.co.uk
Submission type: Invited commentary
Running title: There is an ‘I” in team
Abstract word count: 118
Manuscript word count: 1914
Number of figures: 3
Number of tables: 0
Downloaded by aaron.coutts@uts.edu.au on 07/01/18, Volume ${article.issue.volume}, Article Number ${article.issue.issue}
“Putting the ‘i’ Back in Team” by Ward P, Coutts AJ, Pruna R, McCall A
International Journal of Sports Physiology and Performance
© 2018 Human Kinetics, Inc.
Abstract
There is a common expression in sports, that there is no ‘i’ in team. However, there is also a very
important ‘i’ in sports teams - the individual athlete/player. Each player has his/her own unique
characteristics including physical, physiological and psychological traits. Due to these unique
characteristics, each player requires individual provision - whether it be an injury risk profile and
targeted prevention strategy or treatment/rehabilitation for injury, dietary regimen, recovery or
psychological intervention. The aim of this commentary is to highlight how four high-performance
teams from various professional football codes are analysing individual player data.
Key words: data analysis, individual, team-sports, athletes, monitoring
Downloaded by aaron.coutts@uts.edu.au on 07/01/18, Volume ${article.issue.volume}, Article Number ${article.issue.issue}
“Putting the ‘i’ Back in Team” by Ward P, Coutts AJ, Pruna R, McCall A
International Journal of Sports Physiology and Performance
© 2018 Human Kinetics, Inc.
Introduction
WIthin high-performance sports organisations, a critical question for science/medicine
practitioners is, ‘how best to analyse, interpret and report monitoring and testing protocols in order to
collaborate with the coaching staff on the design and implementation of the training program for both
individual players and the team as a collective.’ The gold standard is likely to follow an evidence-led
approach1,2 using the integration of coaching expertise, athlete values, and the best relevant research
evidence into the decision-making process for the day-to-day service delivery to players”2. The aim of
this commentary is to highlight how an evidence-led approach, targeted at the individual player is
currently being integrated by four high-performance teams from various football codes; Association
Football, American Football and Australian Rules Football.
There is a well-known saying in sport; ‘there is no ‘i’ in team’, and the current authors argue
that, collectively, there is a very important ‘i’ in team – the ‘individual’. In the current authors affiliated
teams, there are collectively > 150 players from > 20 countries spanning 5 continents. With a variety of
ethnicities, each player with unique physiological and psychological characteristics. Each of these
players requires individual catering - whether an injury risk profile, prevention strategy,
treatment/rehabilitation, dietary regimen, recovery or psychological intervention. It is hoped that this
commentary will provide some insights into innovative practices and begin the discussion and sharing
of ideas of how other teams are integrating an evidence-led approach to optimise the servicing of the
individual. ‘The methods of analysis discussed can supplement the daily interaction with players,
allowing recommendations for modifications to training to be discussed with the coaching staff’.
Analysing individual player data
‘In the sport scientist/medicine practitioners’ daily practice, many decisions are made ‘on the fly’
(i.e. within a short timeframe or even immediately).’ For these practitioners, it is important that they have
confidence in their decision making processes before they provide advice to coaches and/or players. As
practitioners, we typically collect a variety of measurements around a players’ physical capacity, wellness,
injury-risk, rehabilitation progress, training/match performance with the intent to identify and monitor the
individual responses and guide specific treatments and/or interventions. A challenge for practitioners is
Downloaded by aaron.coutts@uts.edu.au on 07/01/18, Volume ${article.issue.volume}, Article Number ${article.issue.issue}
“Putting the ‘i’ Back in Team” by Ward P, Coutts AJ, Pruna R, McCall A
International Journal of Sports Physiology and Performance
© 2018 Human Kinetics, Inc.
identifying important trends in player’s data and/or deviations from normal patterns. Whilst monitoring
strategies were first reported for individual sport athletes (e.g., endurance sports) as a way to identify risks of
overtraining3-6, this approach has since been adopted by team-sports to guide the training process2,7. The
information obtained from these measures are commonly used to identify players who are not responding
well to team training, individualising rehabilitation and/or identifying reduced readiness or injury risk7. In
team sports, individualised monitoring can be particularly challenging when working with individuals given
the number of players in a team and the wide variety of training interventions prescribed each day (i.e. many
players full-training, whilst some are prescribed modified programs within the team session and others in
bespoke rehabilitation away from other team members). In these sports the coaching staff usually develop a
training plan based on the team’s technical/tactical strategies required for the upcoming match. In this way,
the risk is that players on the team can be administered similar field-based training programs, often without
sufficient consideration of how each individual may be tolerating the prescribed training. Even for teams
where players are carefully monitored at the individual level, it can still be difficult to individualise the
training program as available approaches to make inferences at the individual level are not well described in
the literature.
A number of statistical approaches have been reported to identify individual differences across
periods of time in physical therapy, exercise, and elite sport8-11. The aim of implementing these analyses is to
enhance the decision-making around issues relating to individual athletes. Analytic approaches such as those
drawing from single-subject research design and time series analysis12, or those using a magnitude-based
inference (MBI) approach to analysing the individual9,13 may provide fast thinking sports science and
medicine practitioners with simple methods of analysing individual player data.
Magnitude-Based Inference for Assessing Individuals
Magnitude-based inference is a statistical approach that allows the practitioner to interpret the
magnitude of the observed effect, relative to some standardized threshold, as being either substantial or
trivial14. While initially proposed for making group comparisons, this method has recently been extended to
assessing individuals9,13,15. Through this approach, practitioners can statistically evaluate the direction of a
Downloaded by aaron.coutts@uts.edu.au on 07/01/18, Volume ${article.issue.volume}, Article Number ${article.issue.issue}
“Putting the ‘i’ Back in Team” by Ward P, Coutts AJ, Pruna R, McCall A
International Journal of Sports Physiology and Performance
© 2018 Human Kinetics, Inc.
player’s trend and the magnitude of that trend over time9, allowing for more meaningful and evidence-led
conversations with coaches regarding the health, well-being and performance capacity of players.
Through the MBI approach, the practitioner can qualitatively assess the observed effect relative to a
smallest worthwhile change (SWC)14. Several methods exist for calculating the SWC such as practical
experience in sport, recent publications identifying typical variations in the variable of interest for a similar
level of player, or more commonly, by multiplying the between-subject (player) standard deviation (SD) by
0.216. While this approach may be useful when analysing performance testing data for groups of athletes (e.g.,
vertical jump for athletes in each position group of an American Football team) the magnitude of between
subject SD falls victim to the heterogeneity of the group15 limiting its utility to detect changes when evaluating
serial measures within the individual (e.g., Daily Wellness Questionnaire data). This has led some authors to
determine the SWC from the within-individual coefficient-of-variation (CV), as this approach is not only
individualized to that person’s typical variation but also takes into account the repeated measures structure of
the data15. For example, Plews and colleagues17 took an individualized MBI approach to evaluate the heart
rate variability (HRV) of elite triathletes during a specific training block. The authors established a SWC in
Ln rMSSD (the natural logarithm of the square root of the mean sum of the squared differences between
adjacent normal R:R intervals) as 0.5*individual athlete CV and found that a large linear decrease in Ln
rMSSD CV for the athlete’s 7-day rolling average revealed a trend towards non-functional over-reaching17.
This approach is similar to examining individualized z-scores, a commonly used method in sports science18.
However, it may be easier for practitioners to explain data to coaches or players/athletes in reference to
percentage changes (e.g. she ran 30% more than usual) rather than explaining data on a standardized scale as
is the case with z-scores. For team sport athletes, using a MBI approach such as we have outlined, would be
appropriate for analysing meaningful changes in training and monitoring data. For example, the training load
and/or wellness response of players on a team can be analysed using a within-individual MBI approach by
applying similar logic to that of Plews and colleagues17. In doing so, the outcomes can be visualized to show
when meaningful changes in the player’s training plan occur. MBI can also be applied to various other team
monitoring aspects. Figure 1 follows the trend of an individual players’ perceived soreness during 20-weeks
Downloaded by aaron.coutts@uts.edu.au on 07/01/18, Volume ${article.issue.volume}, Article Number ${article.issue.issue}
“Putting the ‘i’ Back in Team” by Ward P, Coutts AJ, Pruna R, McCall A
International Journal of Sports Physiology and Performance
© 2018 Human Kinetics, Inc.
of preseason training. This type of visualization has been suggested as a positive feature of taking an MBI
approach to data analysis13.
Statistical Process Control
Statistical Process Control (SPC), sometimes referred to as the “two standard deviation band
method”, is a single-subject approach that can be applied by the practitioner to quickly understand a player’s
trend in any type of monitoring data. Taking its roots in industrial quality control, SPC has since been applied
in both social work and physical therapy settings11,19. The SPC analysis utilizes a control chart, which
visualizes the individual player’s time series data with respect to that individual’s control line (average) and
control limits, often representing 2 SD above and below the mean. Observations, which lie above or below
the upper control limit (UCL) or lower control limit (LCL), are deemed to be “out of statistical control” and
would warrant further investigation. To evaluate whether more subtle shifts in the individual’s trend are taking
place, run time errors can be established to identify, for example, periods of time where several observations
(e.g., 8 or 9) reside above or below the control line indicating a potential shift in the overall process19. The
versatility of SPC provides the practitioner with options of setting their own UCL and LCL. For example,
instead of 2SD control limit, a practitioner may feel it important to be alerted when a value exceeds 1SD8.
The control limits may be initially set with general heuristics in mind (fast working sports science approach),
however, as more data is collected and more detailed analysis is carried out, these should be adjusted to
represent the change that is most meaningful for the given population (slow working sports science approach).
It is common practice for sports practitioners to use the entire historic data of a player/athlete as their
"mean" and "SD", however, this may be limited given the changes in training demands that take place
throughout a season, for example between the pre- and in-season periods20,21. Therefore, it might be
more useful to evaluate data that is recent, for example a 28 day mean and SD or to use some form of
rolling average22.
Figure 2 provides an example of an individual player’s 48 h post-match values from a test of isometric
hamstring force. Amber dots represent when the isometric force values (newtons) drop 1SD below the
player’s mean values and red dots when the force falls beyond 1.5SD. These flags (among other subjective,
Downloaded by aaron.coutts@uts.edu.au on 07/01/18, Volume ${article.issue.volume}, Article Number ${article.issue.issue}
“Putting the ‘i’ Back in Team” by Ward P, Coutts AJ, Pruna R, McCall A
International Journal of Sports Physiology and Performance
© 2018 Human Kinetics, Inc.
objective physical, psychological and contextual factors) are used to inform the decision-making process for
that player’s training program, recovery intervention and physiotherapy treatment prior to the next match.
Mixed Models
One potential limitation of SPC analysis is that several measurements on the individual are
required in order to establish the control line, UCL and LCL. This may be problematic for individuals
who enter the team at different stages of the season (e.g., new players via draft or transfer, trade or free
agent signings) or indeed those with missing data due to poor compliance, technological failure or as
commonly seen when international level players transition from club to national team for major
tournaments. To account for these limitations, while still evaluating individual differences in training,
mixed model analyses have been recommended10,23,24. These types of models allow the practitioner to
leverage the “wisdom of the crowd” via fixed effects while also analysing individual differences
through the specification of random intercepts and/or slopes25. In brief, a mixed-model approach
represents a compromise between pooled data (e.g., the average across all observations) and non-pooled
data (e.g., the averages for the individuals themselves, such as SPC)25. This approach is useful in sports-
science where player’s that have a small number of observations may initially be better represented by
the fixed effects until sufficient observations can be obtained. Similarly, those with a substantial amount
of data will have more individualized slope and regression values around the fixed effects of the model.
The random component of these models allow the practitioner to evaluate how much the individual
athlete deviates from the group. Therefore, such a model may be useful for exploring inter-individual
responses that athletes may have to the prescribed training dose.
Mixed-model approaches have begun to find their way into the sports science literature20,26,27.
However, it is generally uncommon for sports scientists and medicine professionals to discuss random
effects outputs, as most simply interpret the fixed effects portion of the model. Without a more thorough
discussion of the model’s random effects it is challenging to understand the relationship between
individual differences displayed by players. One method to utilize the random effects of the model is to
determine the “effect” of interest based off the between-subject SD14. The typical within-individual
variation can be used to represent the confidence we place on a change in a player’s value from one
Downloaded by aaron.coutts@uts.edu.au on 07/01/18, Volume ${article.issue.volume}, Article Number ${article.issue.issue}
“Putting the ‘i’ Back in Team” by Ward P, Coutts AJ, Pruna R, McCall A
International Journal of Sports Physiology and Performance
© 2018 Human Kinetics, Inc.
measurement to the next. This analytic approach can be visualized with a control chart and error bars
around the observed value. Figure 3 shows the difference between observed and predicted session-RPE
(rating of perceived exertion) training loads for a player group, where the prediction came from a mixed
effects model. When the error bars surrounding the value reside outside of the grey area, then we can
be more confident that the change is meaningful and further investigation is warranted (figure 3).
Take home message
The intention of this commentary to increase the awareness of practitioners who may not be applying
statistical approaches to better understand individual athletes within a team sport setting. We hope that by
implementing these approaches that the fast-thinking practitioners can make confident decisions at the
individual player level. While, we concede that such approaches will require detailed planning and new skill
sets for many sports practitioners, this is the beginning of the push to better understand individual players and
we feel that this is an ideal opportunity to start on the right foot. It is the philosophy and methods of the current
authors’ affiliated teams to adopt such strategies as and when appropriate to both enhance our sports science
and medicine service to our own individual players. We welcome and encourage other sports practitioners
from teams to share other methods of analysing the individual so that we can learn and improve together.
Acknowledgments
The authors would like to acknowledge the sports science, medicine and performance staff of each of the
affiliated teams. In particular, Colin Lewin, Ben Ashworth and Sarah Rudd from Arsenal Football Club.
Downloaded by aaron.coutts@uts.edu.au on 07/01/18, Volume ${article.issue.volume}, Article Number ${article.issue.issue}
“Putting the ‘i’ Back in Team” by Ward P, Coutts AJ, Pruna R, McCall A
International Journal of Sports Physiology and Performance
© 2018 Human Kinetics, Inc.
References
1. Buchheit M. Houston, We Still Have a Problem. Int J Sports Physiol Perform. 2017;12:1111-4.
2. Coutts AJ. Challenges in Developing Evidence-Based Practice in High-Performance Sport. Int J
Sports Physiol Perform. 2017;12:717-8.
3. Foster, C., A. Snyder, and R. Welsh, Monitoring of training, warm up, and performance in athletes,
in Overload, Performance Incompetence and Regeneration in Sport., M. Lehmann, et al.,
Editors. 1999, Kluwer Academic / Plenum Publishers: New York. p. 43-52.
4. Foster, C., et al., Athletic performance in relation to training load. Wisconsin Medical Journal,
1996;95:370-374.
5. Hooper, S.L. and L.T. Mackinnon, Monitoring overtraining in athletes:recommendations. Sports
Medicine, 1995;20:321-327.
6. Rushall, B.S., A tool for measuring stress tolerance in elite athletes. Journal of Applied Sport
Psychology, 1990;9:51-66.
7. Coutts, A.J., S. Crowcroft, and T. Kempton, Developing athlete monitoring systems: Theoretical
basis and practical applications, in Sport, Recovery and Performance: Interdisciplinary
Insights, M. Kellmann and B. Beckmann, Editors. 2017, Routledge: Abingdon. p. in press.
8. Sands WA, Kavanaugh AA, Murray SR, McNeal JR, Jemni M. Modern Techniques and
Technologies Applied to Training and Performance Monitoring. 2017;12:S263-272.
9. Hopkins WG. A spreadsheet for monitoring an individuals changes and trends. Sportsci. 2017;21:5-
9.
10. Kwok OM, Underhill AT, Berry JW, Luo W, Elliot TR, and Yoon M. Analyzing Longitudinal
Data with Multilevel Models: An Example with Individuals Living with Lower Extremity
Intra-articular Fractures. Rehabil Psychol. 2008;53:370-386.
11. Nourbakhsh MR, and Ottenbacher KJ. The Statistical Analysis of Single-Subject Data: A
Comparative Examination. Phys Ther. 1994;74:768-76.
12. Sands WA, McNeal JR, and Stone M. Plaudits and pitfalls in studying elite athletes. Percept Mot
Skills. 2005;1:22-4.
13. Buchheit M. The numbers will love you back in return – I promise. In J Sports Physiol Perf.
2016;11:551-4.
14. Hopkins WG and Batterham AM. Error rates, decisive outcomes and publication bias with several
inferential methods. Sports Med. 2016;46:1563-73.
15. Buchheit M. Monitoring training status with HR measure: do all roads lead to Rome? Front
Physiol. 2014; doi: 10.3389/fphys.2014.00073
16. van Shaik P and Weston M. Magnitude-based inference and its application in user research. Int J
Human-Computer Studies. 2016;88.38-50.
Downloaded by aaron.coutts@uts.edu.au on 07/01/18, Volume ${article.issue.volume}, Article Number ${article.issue.issue}
“Putting the ‘i’ Back in Team” by Ward P, Coutts AJ, Pruna R, McCall A
International Journal of Sports Physiology and Performance
© 2018 Human Kinetics, Inc.
17. Plews DJ, Laursen PB, Kilding AE, Buchheit M. Heart rate variability in elite triathletes, is
variation in variability the key to effective training? A case comparison. Sports Med.
2012;112:3729-41.
18 Robertson. S., Bartlett. JD., Gastin. PB. (2017). Red, Amber, or Green? Athlete monitoring in team
sport: The need for decision-support systems. Int J Sports Physiol Perform. 12(Suppl 2):
S273-S279.
19. Orme JG, and Cox ME. Analyzing single-subject design data using statistical process control
charts. Social Work in Research. 2001;25:115-27.
20. Ritchie. D., Hopkins. WG., Buchheit. M., Courdy. J., Bartlett. JD. (2016). Quantification of
training ad competition load across a season in an elite Australian football club. Int J Sports
Physiol Perform. 11(4): 474-479.
21. Moreiera. A., Bilsborough. JC., Sullivan. CJ., CIancosi. M., Coutts. AJ. (2015). Training
periodization of professional Australian football players during an entire Australian Football
League season. Int J Sports Physiol Perform. 10(5): 566-571.
22. Williams. S., West. S., Cross. MJ., Stokes. KA. (2017). Better way to determine the actute:chronic
workload ratio? Br J Sports Med. 51(3): 209-210.
23. Gastin PG, Meyer D, and Robertson D. Perceptions of wellness to monitor adaptive responses to
training and competition in elite Australian Football. J Strength Cond Res. 2013;27:2518-26.
24. Cnaan A, Laird NM, and Slasor P. Using the general linear mixed model to analyse unbalanced
repeated measures and longitudinal data. Statistics in Medicine. 1997;16:2349-80.
25. Gelman A, Hill J. Data analysis using regression and multilevel/hierarchical models. Cambridge
University Press, New York, NY. 2007.
26. Malone JJ, Di Michele R, Morgans R, Burgess D, Morton JP, Drust B. Seasonal Training Load
Quantification in Elite English Premier League Soccer Players. Int J Sports Physiol Perform.
2015;10:489-97.
27. Kempton T and Coutts AJ. Factors affecting exercise intensity in professional rugby league. J Sci
Med Sport. 2016;19:504-8.
Downloaded by aaron.coutts@uts.edu.au on 07/01/18, Volume ${article.issue.volume}, Article Number ${article.issue.issue}
“Putting the ‘i’ Back in Team” by Ward P, Coutts AJ, Pruna R, McCall A
International Journal of Sports Physiology and Performance
© 2018 Human Kinetics, Inc.
Figure 1. An individual player’s perceived soreness during a 20-week pre-season period. Values that
lie beyond the dashed lines are have a 75% chance of a moderate effect (effect size >0.6).
Downloaded by aaron.coutts@uts.edu.au on 07/01/18, Volume ${article.issue.volume}, Article Number ${article.issue.issue}
“Putting the ‘i’ Back in Team” by Ward P, Coutts AJ, Pruna R, McCall A
International Journal of Sports Physiology and Performance
© 2018 Human Kinetics, Inc.
Figure 2. Statistical Process Control chart displaying Match day + 2 (48h) isometric hamstring force
(Left Hamstring at 30 degrees knee flexion). Amber dot represents 1 standard deviation and Red dot
represents 1.5 standard deviations from player mean values. The dotted lines represent the area where
any scores are considered within statistical control. The practitioner can quickly scan each Match day
+ 2 testing day and see where a players hamstring isometric force is out with their normal range i.e. out-
with statistical control.
Downloaded by aaron.coutts@uts.edu.au on 07/01/18, Volume ${article.issue.volume}, Article Number ${article.issue.issue}
“Putting the ‘i’ Back in Team” by Ward P, Coutts AJ, Pruna R, McCall A
International Journal of Sports Physiology and Performance
© 2018 Human Kinetics, Inc.
Figure 3. Difference between the Observed and Predicted RPE Training Load for players of an
American football team. Predictions were made using a linear effects mixed model. The shaded area
represents a moderate effect relative to the between subject standard deviation. Error bars represent the
within-subject standard deviation from the mixed-model. The practitioner can quickly scan training for
the day and identify any athletes with an abnormal difference between their observed and predicted
responses for this training session.
Downloaded by aaron.coutts@uts.edu.au on 07/01/18, Volume ${article.issue.volume}, Article Number ${article.issue.issue}