Still on the road to Rome

– Written by Mathieu Lacome, Ben M. Simpson and Martin Buchheit, France

PART 1: TRADITIONAL PRACTICES AND NEW CONCEPTS

The assessment of the various training aspects that may help to gain an insight into athletes’ dose-response relationships to training has been the holy grail of coaches and sport scientists for decades¹. The reasons for such interest lie in the need for individualised training to both improve performance and decrease injury risk. Given the high cost of injuries (~US$12.5 million annually per team in the top four football leagues²) and their strong association with team performance³, the interest in both practical and analytical methods that may reduce injuries is warranted.

The last 15 years have seen an incredibly rapid development of (micro)technology in the field⁴. Player tracking has become one of the most important components of load monitoring in team sports⁵. Most professional teams use GPS or alternative tracking systems on a regular basis (e.g. Prozone®, TRACAB®, Inmotio®). Training load reports are now generated within moments following each training session and have become a key element in programming both team and individual training sessions.

This manuscript is the first of a two-part article about training load monitoring in team sports. In this first part, we describe the current and ever-evolving challenges that practitioners face when monitoring their athletes’ training load and health status. We then offer some thoughts on building a framework that may improve current models in an applied setting. The second part provides some guidance on how to improve data visualisation and increase coaching staff ‘buy-in’, which may, in turn, improve their ability to make informed decisions.

CURRENT PRACTICES AND ASSOCIATED CHALLENGES

Since the early 2000s there has been an exponential rise in research about training load monitoring^6-10, allowing sport scientists to base their analyses on strong foundations. While it is not the aim of this paper to build a history of methods, refining current best practices is important to foresee the most relevant future directions.

GPS tracking with basic metrics

Results from a recent survey on the current practices of high-level football clubs to monitor training load showed that of 41 clubs surveyed, 40 collected heart rate and GPS data for every player during every field training session¹¹. Among the top 10 variables used to quantify training load during practice, distance covered in different speed zones, accelerations, heart rate-related variables and accelerometer metrics (e.g. PlayerLoad™¹²) were the most frequently used. Referring to the Gray classification⁵ (Level 1: typical distances covered in different velocity zones; Level 2: all events related to changes in velocity – accelerations, decelerations, and changes of directions; Level 3: all events derived from the inertial sensors/accelerometers), level 1 and 2 type of data are the most used in elite football. Since player activity patterns are more heavily influenced by contextual variables (e.g. rules, coaches’ interventions, scoreline, drills used) than players’ current fitness status¹³, locomotor-related variables (level 1 and 2) may not be the most appropriate to directly monitor players’ training status⁵. Additionally, given the diversity of soccer playing positions and/or player profiles that induce large between-player differences in locomotor activity, comparing locomotor performance between players is not very useful either. It therefore makes more sense to assess changes within individual players. One potential option, although limited, is the use of very standardised drills (such as D-1 game simulations, i.e. 9 vs 9+GKs⁵ [D-1=1 day pre-game, D+1=1 day post-game, etc.]) to assess changes in individual players’ movement strategies (relative to themselves) and gain insight into their training status. However, drill standardisation is not always feasible within the competitive context (e.g. some key players not available to train, congested fixtures minimising access to the drill of interest), which limits this first approach. To compensate for the limitations of level 1 or 2 variables in the absence of standardised drills, sport scientists generally examine a player’s activity using two types of normalisation:

1. Comparing one player’s data over multiples days, using historical data (i.e. intra-player trend).
2. Comparing a player’s data to the rest of the team, with their locomotor activity systematically examined relative to the team or a group of players.

Data analysis

Intra-player analyses

• Daily readiness: with this approach, practitioners monitor changes/trends in individual players’ activity in relation to their own historical data, to track signs of acute fatigue. Comparing activity on a given day with the range of intensity/volume of activity that has been recorded for similar training days (e.g. mean drills response ± standard deviations for this particular player for a D+3 and D-2 session) allows for direct estimations of players’ readiness to perform. However, since differences in the session content (e.g. coach added an extra finishing drill to the usual D-1 session) or context (i.e. playing a Champions League game vs playing the bottom team in the league the next day) may have a larger effect on players’ locomotor activity than changes in their fitness status per se, definitive conclusions remain difficult to draw.
• Acute:chronic ratio: recently, the acute:chronic workload ratio (A/C) has been subject to growing interest as a way to  monitor injury risk^14,15. This model is not aimed at comparing specific session locomotor responses to each other, but rather at tracking the respective changes in the so-called acute (5 to 7 days) and chronic loads (21 to 28 days), using internal and/or external measures of load. Although promising, several limitations prevent elite football clubs from fully utilising this model¹⁶:

1. First, club sport scientists need to:

• Collect enough data (at least 1 full year of training load and injury data) to build an in-house model for the club.
• Find the best A/C ratio split (e.g. 7:28 vs 6:21 days) to fit their sport/club/culture context¹⁷.
• Calculate the optimal ratios for each variable (i.e. sRPE, high-speed running, mechanical work), since there are likely variable-specific sweet spots for decreased injury risk.

2. Second and perhaps most importantly, there must be a way to deal with the frequent periods of international duties of players at elite clubs. For the top clubs in Europe, during international breaks, 60 to 75% of the players are called up to their national teams (either senior or age-group). With little or no data provided to clubs by national teams, club staff would need 21 to 28 days after players return to compute A/C ratios, by which time there may already be another international break (Figure 1). With an international break every month from September to mid-November, the use of A/C ratios is compromised during the first part of the season. While data can be estimated¹⁶, this requires a lot of work and whether it truly reflects players activity remains unclear. Figure 1 shows the data of an international player. During the first part of the season, the three international breaks require practitioners to estimate a significant amount of data (light grey areas), decreasing the confidence in the ratios calculated during this 3-month period. Also, the data can be interpreted differently depending on the windows used (7:28 vs 4:18, sweet spot vs increased injury risk) as seen in the data reported in this figure.
3. Finally, another limitation the authors have noted following 2 years using this method is the increasingly high number of false positives observed each day, especially using exponentially-weighted moving averages¹⁵ (i.e. the ratio is very high but players remain injury-free). This directly discredits sports scientists trying to warn coaches and medical staff of the potential risk of injury to these players. Conversely, the fact that injuries may still occur 1 to 3 weeks¹⁸ following a spike in load helps to offset this apparent error and to justify it as a reliable method for predicting injury. In fact, because of the recurrent spikes in load, once a player is injured it is often possible to find a spike retrospectively! (Figure 1).

These limitations require sport scientists to rely on other methods to analyse data and this is where normalisation of individual players’ locomotor activity relative to the team or a group of players may offer new perspectives.

Between-player normalisation

Player activity normalisation relative to the mean/median of the team or a group of specific players (same profile, same position) is another way to look at the data (see equation below). The main draw of this approach is that it is less likely to be affected by possible changes in session content for a given day (e.g. D+3, D-2 load), since all players complete the same session. Additionally, it can also be used immediately upon return from international break, since the metric does not rely on load data. Once historical data are available (e.g. previous season), insights into players’ fitness and/or early signs of fatigue can be gained while following the trends of these normalised data over several consecutive training days. Sport scientists can, in fact, use any variables (total distance, high-speed distance, accelerations, mechanical work-related variables/minute) to construct similar models. One possibility is the creation of a standardised value for each player, based on the mean or median value of a drill on a given day and known standard deviation of this drill (see equation). For example, following the examination of the variability of most of our drills at PSG (unpublished data), we have chosen to use only game simulations including goalkeepers, possession-based games, and some tactical and technical drills. Warm-ups were removed due to excessive between-player variability and overall low activity volumes. For our model, as the number of players is highly variable and generally less than 20 for the majority of the drills, data is normalised against the median value rather than the mean because it better represents the central tendency in a small population¹⁹.

Individual player readiness = (Drill locomotor value (TD) – drill group median (TDmed)) / (standard deviation of the drill)

However, there are still some limitations with this methodology, especially when:

1. Training with a small group of players (top-up sessions, positional-specific sessions), since the variability of the metrics likely increases with the decreased number of players.
2. A player is returning to play after an injury; as the player trains alone, no comparison can be made (Figure 5, panel Mech W Readiness).

Importance of using the right statistics

The main aim of this paper is not to describe in detail all statistical methods available but to highlight the current best practices to assess changes in monitoring variables in individual players. For more information about the use of statistics in sports sciences, the reader is referred to the recent papers by Buchheit et al^20,21. Figure 2 presents a practical framework for using statistics in applied sport science. In short, data analysis should start by assessing the typical error of measurement (TE) of each metric (or obtaining it from the literature) and defining the smallest worthwhile change (SWC, i.e. what is the smallest meaningful change). Once these two important variables are defined, magnitude-based inferences can be used to compare the change/difference in the variable of interest (±TE) with the SWC and, in turn, provide the staff with meaningful results expressed either in plain wording (e.g. possible small decrease) or simply used to highlight the right numbers.

OPTIMISING THE MODEL

Every practitioner working in elite sport continually looks for improvements in the way they collect, analyse and report data to the coaching staff (Figure 3). To move forward, sport scientists could use a simple, yet effective framework for continual improvement looking at:

1. Adjustment of previously used variables based on contextual variables to decrease noise.
2. Relationships between existing variables.
3. Integration of new and useful variables to try to gain insight into players’ fitness, readiness to perform and/or fatigue (Figure 3).

Adjusting the metrics based on environmental conditions

For outdoor sports, weather and environmental conditions may add noise to measurements and act as additional confounding factors, which should be accounted for when interpreting the data. The following examples show how better insights can be gained using simple statistics.

Monitoring submaximal heart rate (HR) is common today in elite team sports clubs^16,22. Many clubs use 4- to 5-minute steady state (~12 to 14 km/h) running-based monitoring on a weekly/monthly basis¹¹ as an index of cardiovascular fitness. Since HR is closely related to oxygen uptake during continuous exercise, HR during exercise (when expressed as a percentage of maximal HR) provides a good marker of a player’s relative exercise intensity; the lower the HR, the fitter the player^23,24. However, to accordingly assess players training status, practitioners must know both the error of the measurement (3%²⁵) and the magnitude of the changes in HR that matters (Figure 2). Using simple linear relationships between training-induced changes in HR and high-intensity running performance, we previously suggested²⁵ that changes in submaximal heart rate as small as 1% were likely associated with a small but substantial improvement in high-intensity running performance (i.e. SWC (1% for incremental test or 0.2ú between-athletes SD for the 30-15 Intermittent Fitness Test⁵).

However, because of the likely effect of heat on HR responses, adjustments must be made to ensure that the confounding influence of heat on HR is ruled out. One option is to perform the monitoring inside, where the environment can be controlled²⁶. However, in the likelihood of limited available indoor space for running-based monitoring, testing players on exercise bikes is an alternative²⁷, albeit far less engaging for running-based team sport players. On the field, football teams are more inclined to use running-based monitoring, meaning the temperature effect on HR for the given conditions must be known.

Preliminary data collected at PSG have shown that a 10°C increase in temperature leads roughly to a 1% increase in HR% during a 4-minute monitoring run (Figure 4). There can be large temperature variations between summer training camps in hot conditions (temperature >35°C) and the cold winter in France (temperature <0°C). Such variation can lead to changes of up to +/-2% in HR%, which are higher than the SWC and thus meaningful. To avoid misinterpretation (i.e. players are assessed as unfit while the shift in HR% is due to hot temperature), it is necessary to adjust the HR values recorded based on outside temperature (Figure 4, lower panel).

Other changes in environmental conditions could have an impact on recorded metrics. Understanding the influence (or lack of) of the various contextual variables that can affect training load metrics is of great importance. With the recent utilisation of advanced metrics (described later in this paper, level 3 metrics), sport scientists can record contact time, flight time and/or leg stiffness (K)^28,29 during running, which suggests that understanding the player-pitch interaction is of interest. Yet little is known on how pitch surface (hardness, shear strength) can influence these metrics. While preliminary internal studies have shown that slight variation in pitch stiffness as measured with the Clegg hammer (~70 to 85 kN) has no clear influence on metrics related to neuromuscular efficiency during running (K) (Figure 4), future research should examine potential factors that could influence accelerometer data. It is our responsibility as sport scientists to identify every contextual variable that could increase the noise of the model and, in turn, decrease the ability of staff to detect meaningful changes in players’ fitness or fatigue and/or to lead inaccurate interpretations.

Building relationships between existing variables

In football clubs around the world, compound metrics created by combining two or more variables are used, requiring appropriate internal validation. For example, the ratio between velocity load (or metres/min) and force load (or PlayerLoad™) can provide a representation of the amount of ‘force’ or ‘ground impulses’ required per unit of displacement. This metric can be used to assess neuromuscular/running efficiency (the greater the ratio, the better the efficiency) during standardised drills such as box-to-box runs or small-sided games^29,30 (see next section). While this metric still requires proper validation, the concept in itself and the preliminary results are promising.

To assess locomotor/work efficiency and, in turn, infer potential fatigue and readiness to perform outside the laboratory, sport scientists may use internal-to-external load ratios. In other words, simple internal-to-external load ratios can provide insights into players’ training status and may be helpful when deciding whether to alter their training load. Using existing metrics to build relationships between internal and external training load variables is a simple way to assess internal load relative to external load as a cost/output relationship³¹. For example, an increase in internal load relative to a standardised external load (e.g. monitoring run) may infer player fatigue or decreased fitness, while a reduced internal load (e.g. a lower heart rate or perception of effort) during a standardised external load likely indicates that a player is gaining fitness and coping well with training³². Buchheit et al³³ used a ratio between RPE and relative total distance (RPE:m/min) to assess the overall acclimatisation and fatigue trends during a training camp in a hot environment preceded by long-haul flight. In addition, Akubat et al³⁴ observed that the ratio between total distance (TD) and iTRIMP (individualised training impulse, a compound measure based on HR³⁵) – TD:iTRIMP – during a standardised football-specific exercise was related to measures of fitness (velocity at onset of blood lactate accumulation – vOBLA, velocity at lactate threshold – vLT), suggesting that such a ratio could be used as a measure of readiness to perform or running efficiency during official games³⁶.

As recently shown in Australian rules football, measures of internal load (RPE) are related to various external load metrics, while being highly individual³⁷. As such, individual relationships between internal- and external-load-related metrics could be assessed during a pre-determined period to later be used as a prediction model. Then, external load metrics (e.g. GPS) could be used as independent variables to predict the dependent variable internal load (e.g. HR) after sessions. Predicted HR responses could, therefore, be compared to the real HR responses and a HR_real:HR_predicted ratio could, in theory, inform whether the athlete is gaining fitness and coping well with training or becoming fatigued (Figure 5, panel predicted vs real (%)). To sum up the pros and cons of the different methods, Figure 5 presents total distance covered by a player over 2 months, from the start of his rehabilitation post-injury to his full return to play. The different panels present (from top to bottom), total distance covered, and three different options for monitoring injury risk or potential signs of fatigue: acute:chronic ratios, readiness index, and internal:external relationships with HR_predicted vs HR_real comparisons.

Possibility for new variables

With the ever-evolving advances in technology, a new batch of GPS devices will be soon available on the market (GPSport Evo, Statsport APEX, Catapult G5, to name a few) incorporating improved GPS chips (≥15 Hz) and accelerometers (≥400 Hz). Sport scientists will encounter new challenges and opportunities in terms of athlete monitoring. Level 3 (accelerometer) data will be easily available and likely more accurate than in the past.

As previously described⁵, innovative and promising variables will be available for every session and in turn, fatigue monitoring could become much easier and more precise.

• Force load (fL)⁵: with the Athletic Data Innovation analyser (ADI), force load refers to the sum of estimated ground reaction forces during all foot impacts, assessed via the accelerometer-derived magnitude vector. fL reflects only locomotor-related impacts and provides better estimates of overall footwork and impulses than total distance or PlayerLoad™¹², especially when the session includes static movements and low displacement (e.g. rondos, free kicks).
  • During a standardised drill, an average velocity (vL) to force load ratio (vL:fL) can be used to assess neuromuscular/running efficiency (the greater the ratio, the better the efficiency). Recently, the vL:fL ratio during box-to-box runs was shown to decrease following football-specific endurance and speed sessions, suggesting a loss of efficiency in horizontal force application capability (likely due to the fatiguing effect of large amounts of high-speed running or training volume on posterior chain function)²⁹. Also, moderate-to-large increases in vL:fL were observed 2 days after the end of an intense training camp in the heat, suggesting an increase in neuromuscular efficiency, which likely related to a rebound in players’ neuromuscular freshness³⁰.
  • fL can be compared between right and left legs during any locomotive actions (e.g. specifically while accelerating vs running at high speed, which is likely related to the use and potential weaknesses of different muscle groups)²⁸.
• Stride characteristics (contact and flight time calculated from accelerometer data): from these two variables, it is possible to calculate vertical stiffness (K), which has been shown to decrease substantially with neuromuscular fatigue³⁸. While the typical error for K is slightly greater when calculated in the field (i.e. box-to-box runs) compared with standardised runs on an indoor treadmill (11±4.5% vs 6±1.5%^28,29), it remains relatively small. The constant monitoring of stride characteristics during standardised running bouts in the field provides new perspectives for monitoring of neuromuscular status in ecological conditions.

Figure 6 shows how such variables can be used in practice. Panel A shows the symmetry calculated from the fL of all foot impacts when either running above 14.4 km/h, changing direction (CODs) or accelerating (>2 m/s/s) for all the sessions of a player suffering from a tibiofibular inferior ligament sprain (right ankle). Following his injury, there was a clear force deficit on the left side, which progressively returned to baseline as the return-to-play programme advanced. With these novel metrics, especially given that force-load imbalances can be locomotor-phase-specific (i.e. CODs vs accelerations vs high-speed running), detailed patterns can be identified for specific injuries. For example, in panel C, a player with a medial collateral (MCL) sprain in the right knee presented much greater strength-imbalances during CODs phases than during high-speed running and/or accelerations, which is likely due to the specificity of the strain associated with this injury (the MCL is mainly involved in protecting the knee against lateral force and less involved in anteroposterior movements). The diagnosis of strength imbalances is thus locomotion-dependent allowing sport scientists to complement/confirm the doctor or physio’s manual testing. This in turn allows the provision of a fully functional diagnosis of the sprain or imbalance.

CONCLUSION

The role of sport scientists is beginning to be well understood in elite clubs and the peak of inflated expectations has now passed. While many roads lead to Rome, we believe that sport scientists should have a clear vision of the framework required to develop/optimise/improve the models used to analyse training loads. This will help them gain better insight into players’ fitness, readiness to perform and fatigue, and improve the quality and efficiency of their support to coaching staff. But, as we will see in the second part of this manuscript, this is only halfway to Rome and good data visualisation should help practitioners improve coaching staff ‘buy-in’ and, therefore, staff ability to make informed decisions.

References

1. Cardinale M, Varley MC. Wearable training-monitoring technology: applications, challenges, and opportunities. Int J Sports Physiol Perform 2017; 12 (Suppl 2):S255-S262.
2. Abotel K. The Crippling Cost Of Sports Injuries. Forbes 2015. Available from: www.forbes.com/sites/sap/2015/08/11/the-crippling-cost-of-sports-injuries/#7d9eb79b4d1f
3. Hägglund M, Waldén M, Magnusson H, Kristenson K, Bengtsson H, Ekstrand J. Injuries affect team performance negatively in professional football: an 11-year follow-up of the UEFA Champions League injury study. Br J Sports Med 2013; 47:738-742.
4. Malone JJ, Lovell R, Varley MC, Coutts AJ. Unpacking the black box: applications and considerations for using GPS devices in sport. Int J Sports Physiol Perform 2017; 12 (Suppl 2):S218-S226.
5. Buchheit M, Simpson BM. Player tracking technology: half-full or half-empty glass? Int J Sports Physiol Perform 2017; 12 (Suppl 2):S235-S241.
6. Cummins C, Orr R, O'Connor H, West C. Global positioning systems (GPS) and microtechnology sensors in team sports: a systematic review. Sports Med 2013; 43:1025-1042.
7. Halson SL. Monitoring training load to understand fatigue in athletes. Sports Med 2014; 44 (Suppl 2):S139-147.
8. Drew MK, Finch CF. The relationship between training load and injury, illness and soreness: a systematic and literature review. Sports Med 2016; 46:861-883.
9. Aughey RJ. Applications of GPS technologies to field sports. Int J Sports Physiol Perform 2011; 6:295-310.
10. Dellaserra CL, Gao Y, Ransdell L. Use of integrated technology in team sports: a review of opportunities, challenges, and future directions for athletes. J Strength Cond Res 2014; 28:556-573.
11. Akenhead R, Nassis GP. Training load and player monitoring in high-level football: current practice and perceptions. Int J Sports Physiol Perform 2016; 11:587-593.
12. Barrett S, Midgley A, Lovell R. PlayerLoad™: reliability, convergent validity, and influence of unit position during treadmill running. Int J Sports Physiol Perform 2014; 9:945-952.
13. Carling C. Interpreting physical performance in professional soccer match-play: should we be more pragmatic in our approach? Sports Med 2013; 43:655-663.
14. Malone S, Owen A, Newton M, Mendes B, Collins KD, Gabbett TJ. The acute:chonic workload ratio in relation to injury risk in professional soccer. J Sci Med Sport 2017; 20:561-565.
15. Murray NB, Gabbett TJ, Townshend AD, Blanch P. Calculating acute:chronic workload ratios using exponentially weighted moving averages provides a more sensitive indicator of injury likelihood than rolling averages. Br J Sports Med 2017; 51:749-754.
16. Buchheit M. Applying the acute:chronic workload ratio in elite football: worth the effort? Br J Sports Med 2017; 51:1325-1327.
17. Carey DL, Blanch P, Ong KL, Crossley KM, Crow J, Morris ME. Training loads and injury risk in Australian football-differing acute: chronic workload ratios influence match injury risk. Br J Sports Med 2017; 51:1215-1220.
18. Hulin BT, Gabbett TJ, Lawson DW, Caputi P, Sampson JA. The acute:chronic workload ratio predicts injury: high chronic workload may decrease injury risk in elite rugby league players. Br J Sports Med 2016; 50:231-236.
19. Upton G, Cook I. Oxford Dictionary of Statistics (Revised), 2nd ed. Oxford: Oxford University Press 2008.
20. Buchheit M. The numbers will love you back in return-I promise. Int J Sports Physiol Perform 2016; 11:551-554.
21. Buchheit M. Want to see my report, coach? Aspetar Sports Medicine Journal 2017; 6 (Targeted Topic 12):36-43.
22. Thorpe RT, Atkinson G, Drust B, Gregson W. Monitoring fatigue status in elite team-sport athletes: implications for practice. Int J Sports Physiol Perform 2017; 12 (Suppl 2):S227-S234.
23. Buchheit M, Chivot A, Parouty J, Mercier D, Al Haddad H, Laursen PB et al. Monitoring endurance running performance using cardiac parasympathetic function. Eur J Appl Physiol 2010; 108:1153-1167.
24. Mann T, Lamberts RP, Lambert MI. Methods of prescribing relative exercise intensity: physiological and practical considerations. Sports Med 2013; 43:613-625.
25. Buchheit M. Monitoring training status with HR measures: do all roads lead to Rome? Front Physiol 2014; 5:73.
26. Buchheit M, Racinais S, Bilsborough J, Hocking J, Mendez-Villanueva A, Bourdon PC et al. Adding heat to the live-high train-low altitude model: a practical insight from professional football. Br J Sports Med 2013; 47 Suppl 1(Suppl 1):i59-69.
27. Thorpe RT, Strudwick AJ, Buchheit M, Atkinson G, Drust B, Gregson W. Monitoring fatigue during the in-season competitive phase in elite soccer players. Int J Sports Physiol Perform 2015; 10:958-964.
28. Buchheit M, Gray A, Morin JB. Assessing stride variables and vertical stiffness with GPS-embedded accelerometers: preliminary insights for the monitoring of neuromuscular fatigue on the field. J Sports Sci Med 2015; 14:698-701.
29. Buchheit M, Lacome M, Cholley Y, Simpson BM. Neuromuscular responses to conditioned soccer sessions assessed via GPS-embedded accelerometers: insights into tactical periodization. Int J Sports Physiol Perform 2018; 13:577-583.
30. Racinais S, Buchheit M, Bilsborough J, Bourdon PC, Cordy J, Coutts AJ. Physiological and performance responses to a training camp in the heat in professional Australian football players. Int J Sports Physiol Perform 2014; 9:598-603.
31. Burgess DJ. Application of Monitoring Research in Elite Team Sport Periodisation. Available from: https://vimeo.com/159904162
32. Bourdon PC, Cardinale M, Murray A, Gastin P, Kellmann M, Varley MC et al. Monitoring athlete training loads: consensus statement. Int J Sports Physiol Perform 2017; 12 (Suppl 2):S2161-S2170.
33. Buchheit M, Cholley Y, Lambert P. Psychometric and physiological responses to a preseason competitive camp in the heat with a 6-hour time difference in elite soccer players. Int J Sports Physiol Perform 2016; 11:176-181.
34. Akubat I, Barrett S, Abt G. Integrating the internal and external training loads in soccer. Int J Sports Physiol Perform 2014; 9:457-462.
35. Akubat I, Patel E, Barrett S, Abt G. Methods of monitoring the training and match load and their relationship to changes in fitness in professional youth soccer players. J Sports Sci 2012; 30:1473-1480.
36.  Torreño N, Munguía-Izquierdo D, Coutts A, Villarreal ES, Asian-Clemente J, Suarez-Arrones L. Relationship between external and internal loads of professional soccer players during full matches in official games using global positioning systems and heart-rate technology. Int J Sports Physiol Perform 2016; 11:940-946.
37. Bartlett JD, O'Connor F, Pitchford N, Torres-Ronda L, Robertson SJ. Relationships between internal and external training load in team-sport athletes: evidence for an individualized approach. Int J Sports Physiol Perform 2017; 12:230-234.
38. Morin JB, Jeannin T, Chevallier B, Belli A. Spring-mass model characteristics during sprint running: correlation with performance and fatigue-induced changes. Int J Sports Med 2006; 27:158-165.

Mathieu Lacome Ph.D.

Sports Scientist

Ben M. Simpson M.Sc.

Sports Scientist

Martin Buchheit Ph.D.

Head of Performance

Paris Saint-Germain Football Club

Paris, France

Contact: mlacome@psg.fr