Accelerometry-based variables in professional soccer players: comparisons between periods of the season and playing positions.

The aim of this study was to provide reference data of variation in external training loads for weekly periods within the annual season. Specifically, we aimed to compare the weekly acute load, monotony, and training strain of accelerometry-based measures across a professional soccer season (pre-season, first and second halves of the season) according to players' positions. Nineteen professional players were monitored daily for 45 weeks using an 18-Hz global positioning system to obtain measures of high metabolic load distance (HMLD), impacts, and high intensity accelerations and decelerations. Workload indices of acute load, training monotony, and training strain were calculated weekly for each of the measures. The HMLD had greater training strain values in the pre-season than in the first (p ≤ 0.001; d = 0.793) and second halves of the season (p ≤ 0.001; d = 0.858). Comparisons between playing positions showed that midfielders had the highest weekly acute load of HMLD (6901 arbitrary units [AU]), while central defenders had the lowest (4986 AU). The pre-season period was associated with the highest acute and strain load of HMLD and number of impacts, with a progressive decrease seen during the season. In conclusion, coaches should consider paying greater attention to variations in HMLD and impacts between periods of the season and between players to individualize training accordingly.

INTRODUCTION

The individualization of the training process requires the systematic monitoring of training’s impact on players, namely by controlling training load and well-being measures of readiness [1]. Among these aspects, training load quantification has become an important area of study within sports sciences departments [2]. Load monitoring allows the collection of either objective or subjective measures that provide information that is useful to understand the dynamics of load and promoting adjustments in the training process [3]. Two main dimensions are included in load monitoring [4]: (i) external load, which is associated with the physical demands or mechanical work promoted by the exercise, and (ii) internal load, which is associated with psychobiological responses to external load. Usually, in team sports, external load is quantified by positioning-derived data (e.g., data obtained from global positioning systems [GPSs] or multi-camera tracking systems) or accelerometry-derived data (e.g., data obtained from inertial sensor units [IMUs]; or accelerometers) [5]. However, GPS units are commonly equipped with accelerometers or IMUs, which makes it possible to obtain position-derived data (e.g., distances covered at different speeds, changes in velocity) and accelerometry-derived data (e.g., accelerations/decelerations, player load, impacts, stride) at the same time [6, 7]. The integration of accelerometers allows researchers to continuously record signals at a high measuring frequency (commonly, 100 Hz), thus making it possible to acquire summative measures such as player load, dynamic stress load, or metabolic power [8, 9]. However, there are some issues related to the use and interpretation of measures derived from GPSs [6, 10]. Some measures (e.g., distance) can be highly dependent on tactical contexts and players’ fitness status, among other factors [6]. On the other hand, accelerometry-derived measures seem to be more stable even considering variations across sessions and matches, and so they are better for monitoring specific elements, such as fitness or fatigue, over time [6]. Training load monitoring might recognize some training principles such as individualization, progression, overload, or variability of the stimulus [11]. Additionally, within- and between-week variations in external load can be monitored to reduce the chances of overtraining and undertraining [12, 13].

Some workload measures have been used to monitor training load. Among others, the acute-to-chronic workload ratio (acute load divided by chronic load) [14] could control weekly progression and overload in players. However, other important aspects related to the capacity to identify exposures to high doses and minimal within-week variability can be controlled using training monotony and training strain [15]. These two measures were introduced to monitor exposure to bad overreaching or overtraining using the session rate of perceived exertion (sRPE), which is RPE multiplied by the duration of a training session, in minutes [15]. Training monotony is the daily mean load divided by the week standard deviation load, while training strain is the product of weekly training load and monotony [15]. These concepts have been tested to identify possible exposures to injury risk or illness, or associations with decrements in performance [16-19]. Despite attempts to use sRPE to control overload, some evidence suggests that training strain and monotony may better reflect players’ exposure to injury risk [17, 20]. However, the relationships of training monotony and training strain with exposure to injury or illness risk are likely complex [21]. In this sense, modulators such as players’ fitness (i.e., professional vs. amateur) and the season period (i.e., pre-season vs. in-season) must be analysed. Although these specific definitions of monotony and strain are restrictive and likely do not reflect the complexity involved in defining actual monotony and strain, we use these well-established concepts in the present work. Despite the regular practice of controlling training monotony and strain using sRPE, these indices have not been used often to assess external load. One of the few studies that has done this [22] was employed using a new measure that integrates internal and external load measures. However, mechanical work resulting from training sessions should also be considered as a factor that could induce adaptations or fatigue. Furthermore, controlling within-week variations and changes across the weeks could provide sports scientists and coaches with important information about possible differences at different moments of the season or even between playing positions. Based on previous studies [23], meaningful variations in weekly loads can be found between different moments within a season. Moreover, considering the great number of drills based on the game, it is also expected that meaningful variations in external load could occur between playing positions. Therefore, the aim of this study was to describe and compare the weekly acute load, monotony, and training strain of accelerometry-based measures across different moments of a professional soccer season (pre-season, first and second halves of the season) according to players’ positions. It is hypothesized that greater workload indices occur during the pre-season and that meaningful variations occur in the workload indices between playing positions.

MATERIALS AND METHODS

Experimental approach and procedures

Using a descriptive research design, a squad of 19 professional male soccer players was monitored daily throughout a full season. The external load monitored by an 18-Hz GPS unit (including accelerometer) was controlled in all training sessions (n = 197) and matches (n = 44; including national league, national cup, and European league matches). Data covering 45 weeks were analysed. The study period began at the beginning of the pre-season (July 3, 2018) and lasted until the end of the season (May 9, 2019). The season was organized into three periods: (i) pre-season (PS: week 1 to week 6, no official matches); (ii) first half of the season (1stHS: week 6 to week 33, covering the period from the first to the last match of the first round of the national league); and (iii) the second half of the season (2ndHS: week 34 to week 45, covering the period from the first to the last match of the second round of the national league). A graphical depiction of the weekly distribution of training sessions and matches across the season is presented in Figure 1.

FIG. 1

Weekly distribution of training sessions and matches across the season. w: week

The following accelerometry-derived measures were monitored daily for each player: (i) high metabolic power distance; (ii) number of impacts; (iii) high intensity accelerations and decelerations. These measures were then calculated weekly to obtain values for acute load, training monotony, and training strain for each player.

Participants

Nineteen elite professional male players (age: 26.5 ± 4.3 years; body mass: 75.6 ± 9.6 kg; height: 180.2 ± 7.3 cm; experience as professionals: 7.5 ± 4.3 years) from a European First League team participated in this study. Players were categorized by playing position as external defenders (ED, n = 3), central defenders (CD, n = 4), midfielders (MF, n = 6), wingers (W, n = 4), and strikers (ST, n = 2). Inclusion criteria were (i) the player must belong to the team from day 1 to the last day of the season; (ii) players could not stop training for more than two consecutive weeks (due to injuries or illness); (iii) players participated in more than 80% of the training sessions. From a total of 31 players, nine were excluded based on these criteria. The remaining three were excluded for acting as goalkeepers. The players were familiarized with the study design and protocol, as well as with the daily procedures before beginning. After their agreement, they signed a free informed consent form. The study followed the ethical guidelines of the Declaration of Helsinki. The study was approved by the scientific council of Escola Superior de Desporto e Lazer (Portugal).

External load monitoring

External load was quantified using an 18-Hz GPS with a 100-Hz gyroscope, 100-Hz tri-axial accelerometer, and 10-Hz magnetometer (STATSports, Apex, Northern Ireland). This GPS unit was previously tested for validity and reliability, with good levels of accuracy (< 2.3% coefficient of variation) [24] and excellent levels of inter-unit reliability for peak running velocity observed [25]. The number of satellites during data collection ranged between 17 and 21. Each player used the same unit during the period of data collection to reduce possible inter-unit variability. Each player used a vest in which the GPS unit was positioned between their scapulae. The data obtained from the GPS were downloaded and analysed in specific software (STATSports Apex software, version 5.0).

The following measures were collected daily from each player: (i) high metabolic load distance (HMLD: corresponding to the distance covered at a speed of > 5.5 m/s-1 while accelerating/decelerating at ≥ 2 m/s-2); (ii) impacts (Imp: the number of impacts, which are considered instantaneous moments throughout a training session measured in G-forces and expressed as quantity); (iii) high intensity accelerations and decelerations (HA and HD: number of accelerations and decelerations at ≥ 3 m/s-2 maintained for ≥0.5 seconds). Weekly acute load (within-week training sessions and matches summed load), training monotony (mean of training load during the seven days of the week divided by its standard deviation) and training strain (sum of training load for all training sessions and matches during a week multiplied by training monotony) were calculated for each variable and for each player. Considering the acute load, training monotony and training strain (weekly representation), the GPS measures were calculated as follows: (i) weekly HMLD (wHMLD); (ii) mHMLD (monotony HMLD); (iii) sHMLD (strain HMLD); (iv) wImp (weekly Imp); (v) mImp (monotony Imp); (vi) sImp (strain Imp); (vii) wHA (weekly HA); (viii) mHA (monotony HA); (ix) sHA (strain HA); (x) wHD (weekly HD); (xi) mHD (monotony HD); and (xii) sHD (strain HD).

Statistical procedures

Means (with standard deviation) are indicated. Normality (N > 30, thus assuming the central limit theorem) and homogeneity (Levene; p > 0.05) were preliminarily tested and confirmed. The weekly load (acute; monotony; strain) was compared between periods of the season (PS; 1stHS; 2ndHS) using a repeated measures ANOVA followed by Tukey’s HSD post hoc test for pairwise comparisons. To compare playing positions (ED; CD; MF; W; ST), a one-way ANOVA was used, followed by the Tukey HSD post hoc test for pairwise comparisons. Both tests were executed in SPSS software (version 25.0, IBM, Chicago, USA), with p < 0.05. The magnitudes of differences in pairwise comparisons were tested using the standardized effect size of Cohen (d) for a 95% confidence interval (95%CI). The inference of magnitudes was made using the following thresholds [26]: [0.0;0.2], trivial; [0.2;0.6], small; [0.6;1.2], moderate; [1.2; 2.0], large; > 2.0, very large.

RESULTS

The weekly changes in acute load and training monotony over the season for HMLD can be found in Figure 2 (a). The highest weekly acute load (12,277 m) was reached in week 1, and the lowest (2942 m) was recorded in week 43. The highest weekly acute load increase was 99% (from week 29 to week 30), and the largest decrease was -63% (week 19 to 20). Training monotony was the highest in week 45 (5.7 arbitrary units [AU]) and the lowest in week 24 (0.7 AU). The greatest between-week increase (438%) in training monotony occurred between weeks 44 and 45, while the largest decrease (-61%) occurred between weeks 10 and 11. Training strain was highest in week 1 (31638 AU) and lowest in week 44 (2912 AU), while the largest increase (265%) was found from week 32 to week 33, and the largest decrease (-63%) was found from week 42 to week 43 (Figure 2 (b)).

FIG. 2

Weekly distribution across the season for (a) weekly high metabolic load distances (wHMLD), monotony HMLD (mHMLD), and (b) strain HMLD (sHMLD).

Figure 3 (a) shows the weekly changes in acute load and training monotony for high intensity accelerations. The highest weekly acute load reached 1833 n in week 16, and the lowest load was found in week 10 (384 n). The greatest increase (179%) occurred between week 29 and week 30, while the largest decrease (-72%) occurred between week 37 and week 38. Training monotony was highest in week 38 (5.2 AU) and lowest in week 10 (0.7 AU). The largest increase, reaching 311%, was found from week 32 to 33, while the largest decrease (-77%) was found from week 38 to week 39. As seen in Figure 3 (b), training strain was highest in week 1 (4781 AU) and lowest in week 10 (415 AU). The largest increase (254%) was found from week 32 to week 33, while the largest decrease (-74%) was found from week 34 to week 35.

FIG. 3

Descriptive statistics of (a) acute load and training monotony and (b) training strain for high intensity accelerations.

Figure 4 (a) shows the weekly changes of acute load and training monotony for high intensity decelerations. The highest weekly acute load (1762 n) was reached in week 2, while the lowest (418 n) was reached in week 20. The largest increase (158%) was observed from week 15 to week 16, and the largest decrease (-61%) occurred from week 19 to week 20. Training monotony was highest in week 33 (5.2 AU) and lowest in week 10 (0.7 AU). The largest increase in training monotony (382%) occurred from week 32 to week 33, while the largest decrease (-76%) was observed from week 38 to week 39. Figure 4 (b) indicates that training strain was highest in week 33 (4909 AU) and lowest in week 10 (341 AU). The largest increase in training strain (443%) was reached from week 32 to week 33, while the largest decrease (-56%) was found from week 34 to week 35.

FIG. 4

Descriptive statistics of (a) acute load and training monotony and (b) training strain for high intensity decelerations.

Figure 5 (a) shows the weekly changes in acute load and training monotony for impacts. The highest weekly acute load (3333 n) was reached in week 2, and the lowest acute load was found in week 31 (336 n). The largest increase in acute load (194%) occurred from week 15 to week 17, and the largest decrease (-78%) occurred from week 30 to week 31. Training monotony was highest in week 1 (3.8 AU) and lowest in weeks 9 and 24 (0.6 AU). The largest increase in training monotony (404%) was found from week 9 to week 10. As seen in Figure 5 (b), training strain was highest in week 1 (12,949 AU) and lowest in week 27 (388 AU). The largest increase in training strain (495%) occurred from week 32 to week 33, while the largest decrease (-54%) was found from week 5 to week 6.

FIG. 5

Descriptive statistics of (a) acute load and training monotony and (b) training strain for impacts.

Table 1 presents the differences between the PS, 1stHS, and 2ndHS for AL, TM, and TS for HMLD, Imp, HA, and HD. To simplify the description, only moderate to large ESs will be described here. In relation to wHMLD, meaningfully greater TS values were observed in the PS than in the 1stHS (88%) and 2ndHS (46%). Also, wImp was meaningfully greater in the PS than in the 1stHS (74%) and 2ndHS (66%). Similarly, mImp was meaningfully greater in PS than in the 1stHS (50%) and 2ndHS (50%). Moreover, sImp was meaningfully greater in the PS than in the 1stHS (167%) and 2nHS (145%).

TABLE 1

Descriptive statistics (mean ± SD) of acute load, training monotony, and training strain for external load measures in the pre-season, first half of the season, and second half of the season.

	PS (mean ± SD)	1stHS (mean ± SD)	2ndHS (mean ± SD)	p	ES
wHMLD (m)	7875.0 ± 6007.1	6317.5 ± 4966.9	5384.7 ± 4508.8	PS vs. 1stHS: 0.027*PS vs. 2ndHS: ≤ 0.001*1stHF vs. 2ndHS: 0.05	PS vs. 1stHS: 0.304 smallPS vs. 2ndHS: 0.504 small1stHF vs. 2ndHS: 0.193 trivial
mHMLD (AU)	2.0 ± 1.2	1.5 ± 1.1	1.5 ± 1.0	PS vs. 1stHS: ≤ 0.001*PS vs. 2ndHS: ≤ 0.001*1stHF vs. 2ndHS: 0.990	PS vs. 1stHS: 0.449 smallPS vs. 2ndHS: 0.474 small1stHF vs. 2ndHS: 0.000 trivial
sHMLD (AU)	16360.0 ± 14361.3	8717.9 ± 8719.5	7520.7 ± 8590.3	PS vs. 1stHS: ≤ 0.001*PS vs. 2ndHS: ≤ 0.001*1stHF vs. 2ndHS: 0.265	PS vs. 1stHS: 0.793 moderatePS vs. 2ndHS: 0.858 moderate1stHF vs. 2ndHS: 0.138 trivial
wImp (n)	2190.7 ± 1695.7	1258.1 ± 1175.3	1314.4 ± 1109.9	PS vs. 1stHS: ≤ 0.001*PS vs. 2ndHS: ≤ 0.001*1stHF vs. 2ndHS: 0.835	PS vs. 1stHS: 0.742 moderatePS vs. 2ndHS: 0.683 moderate1stHF vs. 2ndHS: 0.049 trivial
mImp (AU)	1.8 ± 1.0	1.2 ± 0.9	1.2 ± 0.8	PS vs. 1stHS: ≤ 0.001*PS vs. 2ndHS: ≤ 0.001*1stHF vs. 2ndHS: 0.740	PS vs. 1stHS: 0.656 moderatePS vs. 2ndHS: 0.701 moderate1stHF vs. 2ndHS: 0.000 trivial
sImp (AU)	4450.3 ± 4482.7	1666.5 ± 2125.4	1815.4 ± 2062.7	PS vs. 1stHS: ≤ 0.001*PS vs. 2ndHS: ≤ 0.001*1stHF vs. 2ndHS: 0.733	PS vs. 1stHS: 1.092 moderatePS vs. 2ndHS: 0.930 moderate1stHF vs. 2ndHS: -0.070 trivial
wHA (n)	1137.2 ± 828.5	985.3 ± 707.2	908.4 ± 645.1	PS vs. 1stHS: 0.181PS vs. 2ndHS: 0.036*1stHF vs. 2ndHS: 0.367	PS vs. 1stHS: 0.210 smallPS vs. 2ndHS: 0.329 small1stHF vs. 2ndHS: 0.112 small
mHA (AU)	1.7 ± 0.8	1.7 ± 0.9	1.8 ± 1.0	PS vs. 1stHS: >0.999PS vs. 2ndHS: 0.6171stHF vs. 2ndHS: 0.294	PS vs. 1stHS: 0.000 trivialPS vs. 2ndHS: -0.105 trivial1stHF vs. 2ndHS: -0.107 trivial
sHA (AU)	2222.8 ± 2162.9	1974.4 ± 1794.6	1808.3 ± 1514.2	PS vs. 1stHS: 0.489PS vs. 2ndHS: 0.1831stHF vs. 2ndHS: 0.487	PS vs. 1stHS: 0.134 trivialPS vs. 2ndHS: 0.243 small1stHF vs. 2ndHS: 0.097 trivial
wHD (AU)	1156.5 ± 773.5	824.1 ± 610.1	817.1 ± 588.3	PS vs. 1stHS: ≤ 0.001*PS vs. 2ndHS: ≤ 0.001*1stHF vs. 2ndHS: 0.989	PS vs. 1stHS: 0.523 smallPS vs. 2ndHS: 0.529 small1stHF vs. 2ndHS: 0.012 trivial
mHD (AU)	1.7 ± 0.8	1.6 ± 1.1	1.7 ± 0.9	PS vs. 1stHS: 0.575PS vs. 2ndHS: >0.9991stHF vs. 2ndHS: 0.279	PS vs. 1stHS: 0.094 trivialPS vs. 2ndHS: 0.000 trivial1stHF vs. 2ndHS: -0.096 trivial
sHD (AU)	1810.4 ± 1760.4	1465.2 ± 1442.5	1455.5 ± 1449.2	PS vs. 1stHS: 0.149PS vs. 2ndHS: 0.1751stHF vs. 2ndHS: 0.996	PS vs. 1stHS: 0.232 smallPS vs. 2ndHS: 0.232 small1stHF vs. 2ndHS: 0.007 trivial

HA: high intensity accelerations; HD: high intensity decelerations; HMLD: high metabolic load distance; wHMLD: weekly HMLD; mHMLD: monotony HMLD; sHMLD: strain HMLD; wImp: weekly impacts; mImp: monotony impacts; sImp: strain impacts; wHA: weekly HA; mHA: monotony HA; sHA: strain HA ; wHD: weekly HD; mHD: monotony HD; sHD: strain HD; PS: pre-season period; 1stHS: first half of the season; 2ndHS: second half of the season; AU: arbitrary units; ES: effect size.

Tables 2, 3, 4, and 5 present the differences between playing positions, AL, TM, and TS for HMLD, Imp, HA, and HD. To simplify the description, only small ESs will be described here. Significantly greater wHMLD were found for ED, MF, W, ST than for CD (24, 38, 31%, and 17%, respectively). Also, MF and W had significantly greater wHMLD values than ST (18% and 12%, respectively). No significant differences were found between positions for mHMLD, while for sHMLD greater values were found for W than for CD and ST (30% and 26%, respectively).

TABLE 2

Descriptive statistics (mean ± SD) of acute load, training monotony, and training strain for HMLD between playing positions.

	ED (mean ± SD)	CD (mean ± SD)	MF (mean ± SD)	W (mean ± SD)	ST (mean ± SD)	p	ES
wHMLD (m)	6194.8 ± 5229.2	4985.5 ± 3915.4	6900.5 ± 5414.9	6514.2 ± 5070.5	5840.9 ± 4521.3	ED vs. CD: 0.204ED vs. MF: 0.632ED vs. W: 0.978ED vs. ST: 0.985CD vs. MF: 0.002*CD vs. W: 0.056CD vs. ST: 0.725MF vs. W: 0.942MF vs. ST: 0.462W vs. ST: 0.858	ED vs. CD: 0.260 smallED vs. MF: -0.132 trivialED vs. W: -0.062 trivialED vs. ST: 0.070 trivialCD vs. MF: -0.392 smallCD vs. W: -0.336 smallCD vs. ST: -0.207 smallMF vs. W: 0.073 trivialMF vs. ST: 0.204 smallW vs. ST: 0.138 small
mHMLD (AU)	1.5 ± 0.9	1.6 ± 1.3	1.5 ± ±	1.5 ± 1.1	1.4 ± 0.6	ED vs. CD: 0.743ED vs. MF: >0.999ED vs. W: 0.996ED vs. ST: 0.973CD vs. MF: 0.631CD vs. W: 0.913CD vs. ST: 0.491MF vs. W: 0.989MF vs. ST: 0.977W vs. ST: 0.889	ED vs. CD: 0.090 trivialED vs. MF: 0.000 trivialED vs. W: 0.000 trivialED vs. ST: 0.123 trivialCD vs. MF: 0.081 trivialCD vs. W: 0.083 trivialCD vs. ST: 0.182 trivialMF vs. W: 0.000 trivialMF vs. ST: 0.093 trivialW vs. ST: 0.104 trivial
sHMLD (AU)	9355.9 ± 10346.4	7910.8 ± 7884.9	9127.3 ± 10137.4	10313.6 ± 10433.2	8202.4 ± 7763.4	ED vs. CD: 0.698ED vs. MF: 0.999ED vs. W: 0.907ED vs. ST: 0.910CD vs. MF: 0.763CD vs. W: 0.202CD vs. ST: >0.999MF vs. W: 0.763MF vs. ST: 0.948W vs. ST: 0.508	ED vs. CD: 0.156 trivialED vs. MF: 0.022 trivialED vs. W: -0.092 trivialED vs. ST: 0.121 trivialCD vs. MF: -0.130 trivialCD vs. W: -0.258 smallCD vs. ST: -0.037 trivialMF vs. W: -0.116 trivialMF vs. ST: 0.097 trivialW vs. ST: 0.220 small

wHMLD: weekly high metabolic load distance; mHMLD: monotony high metabolic load distance; sHMLD: strain high metabolic load distance; ED: external defender; CD: central defender; MF: midfielder; W: winger; ST: striker; AU: arbitrary units; ES: effect size

TABLE 3

Descriptive statistics (mean ± SD) of acute load, training monotony, and training strain for the impacts between playing positions.

	ED (mean ± SD)	CD (mean ± SD)	MF (mean ± SD)	W (mean ± SD)	ST (mean ± SD)	p	ES
wImp (n)	1569.4 ± 1490.0	1199.6 ± 963.9	1617.6 ± 1325.4	1099.1 ± 1086.2	1044.4 ± 1000.5	ED vs. CD: 0.064ED vs. MF: 0.995ED vs. W: 0.006*ED vs. ST: 0.016*CD vs. MF: 0.011*CD vs. W: 0.952CD vs. ST: 0.892MF vs. W: ≤0.001*MF vs. ST: 0.003*W vs. ST: 0.998	ED vs. CD: 0.292 smallED vs. MF: -0.035 trivialED vs. W: 0.361 smallED vs. ST: 0.390 smallCD vs. MF: -0.349 smallCD vs. W: 0.098 trivialCD vs. ST: 0.159 trivialMF vs. W: 0.420 smallMF vs. ST: 0.458 smallW vs. ST: 0.052 trivial
mImp(AU)	1.3 ± 0.9	1.3 ± 0.8	1.3 ± 1.1	1.2 ± 1.0	1.1 ± 0.7	ED vs. CD: 0.990ED vs. MF: >0.999ED vs. W: 0.537ED vs. ST: 0.642CD vs. MF: 0.970CD vs. W: 0.843CD vs. ST: 0.872MF vs. W: 0.373MF vs. ST: 0.526W vs. ST: >0.999	ED vs. CD: ≤0.001 trivialED vs. MF: ≤ 0.001 trivialED vs. W: 0.105 trivialED vs. ST: 0.239 smallCD vs. MF: ≤0.001 trivialCD vs. W: 0.115 trivialCD vs. ST: 0.286 smallMF vs. W: 0.094 trivialMF vs. ST: 0.198 trivialW vs. ST: 0.110 trivial
sImp (AU)	2694.7 ± 3514.1	1874.1 ± 2350.6	2116.7 ± 2399.2	1382.5 ± 1893.9	1411.1 ± 1565.2	ED vs. CD: 0.040*ED vs. MF: 0.178ED vs. W: ≤ 0.001*ED vs. ST: 0.002*CD vs. MF: 0.897CD vs. W: 0.440CD vs. ST: 0.685MF vs. W: 0.040*MF vs. ST: 0.204W vs. ST: >0.999	ED vs. CD: 0.272 smallED vs. MF: 0.199 trivialED vs. W: 0.466 smallED vs. ST: 0.427 smallCD vs. MF: -0.102 trivialCD vs. W: 0.232 smallCD vs. ST: 0.220 smallMF vs. W: 0.333 smallMF vs. ST: 0.319 smallW vs. ST: -0.016 trivial

wImp: weekly impacts; mImp: monotony impacts; sImp: strain impacts; ED: external defender; CD: central defender; MF: midfielder; W: winger; ST: striker; AU: arbitrary units; ES: effect size

TABLE 4

Descriptive statistics (mean ± SD) of acute load, training monotony, and training strain for the HA between playing positions.

	ED (mean ± SD)	CD (mean ± SD)	MF (mean ± SD)	W (mean ± SD)	ST (mean ± SD)	p	ES
wHA (n)	1064.8 ± 742.2	1035.2 ± 699.0	911.8 ± 718.7	995.4 ± 697.0	941.8 ± 600.9	ED vs. CD: 0.996ED vs. MF: 0.210ED vs. W: 0.632ED vs. ST: 0.702CD vs. MF: 0.456CD vs. W: 0.859CD vs. ST: 0.874MF vs. W: 0.975MF vs. ST: 0.997W vs. ST: <0.999	ED vs. CD: 0.041 trivialED vs. MF: 0.210 smallED vs. W: 0.152 trivialED vs. ST: 0.176 trivialCD vs. MF: 0.174 trivialCD vs. W: 0.114 trivialCD vs. ST: 0.140 trivialMF vs. W: -0.061 trivialMF vs. ST: -0.043 trivialW vs. ST: 0.020 trivial
mHA (AU)	1.8 ± 1.0	1.8 ± 0.9	1.7 ± 0.9	1.8 ± 1.0	1.8 ± ±	ED vs. CD: 0.996ED vs. MF: 0.860ED vs. W: 0.969ED vs. ST: 0.989CD vs. MF: 0.653CD vs. W: 0.999CD vs. ST: <0.999MF vs. W: 0.451MF vs. ST: 0.698W vs. ST: <0.999	ED vs. CD: ≤0.001 trivialED vs. MF: 0.106 trivialED vs. W: ≤0.001 trivialED vs. ST: ≤0.001 trivialCD vs. MF: 0.111 trivialCD vs. W: ≤0.001 trivialCD vs. ST: ≤0.001 trivialMF vs. W: -0.106 trivialMF vs. ST: -0.114 trivialW vs. ST: ≤0.001 trivial
sHA (AU)	1977.4 ± 1782.3	2075.7 ± 1738.1	1807.2 ± 1666.3	2148.1 ± 1934.0	1717.2 ± 1669.3	ED vs. CD: 0.989ED vs. MF: 0.913ED vs. W: 0.822ED vs. ST: 0.989CD vs. MF: 0.610CD vs. W: 0.997CD vs. ST: 0.595MF vs. W: 0.339MF vs. ST: 0.995W vs. ST: 0.392	ED vs. CD: -0.056 trivialED vs. MF: 0.099 trivialED vs. W: -0.092 trivialED vs. ST: 0.149 trivialCD vs. MF: 0.159 trivialCD vs. W: -0.039 trivialCD vs. ST: 0.209 smallMF vs. W: -0.192 trivialMF vs. ST: 0.054 trivialW vs. ST: 0.233 small

wHA: weekly high intensity accelerations; mHA: monotony high intensity accelerations; sHA: strain high intensity accelerations; ED: external defender; CD: central defender; MF: midfielder; W: winger; ST: striker; AU: arbitrary units; ES: effect size

TABLE 5

Descriptive statistics (mean ± SD) of acute load, training monotony, and training strain for the HD between playing positions.

	ED (mean ± SD)	CD (mean ± SD)	MF (mean ± SD)	W (mean ± SD)	ST (mean ± SD)	p	ES
wHD (n)	956.8 ± 637.3	786.0 ± 595.0	820.0 ± 650.8	917.8 ± 650.9	755.5 ± 535.8	ED vs. CD: 0.118ED vs. MF: 0.206ED vs. W:0.981ED vs. ST: 0.127CD vs. MF: 0.986CD vs. W: 0.351CD vs. ST: 0.997MF vs. W: 0.550MF vs. ST: 0.931W vs. ST: 0.319	ED vs. CD: 0.277 smallED vs. MF: 0.212 smallED vs. W: 0.060 trivialED vs. ST: 0.333 smallCD vs. MF: -0.054 trivialCD vs. W: -0.211 smallCD vs. ST: 0.053 trivialMF vs. W: -0.150 trivialMF vs. ST: 0.104 trivialW vs. ST: 0.264 small
mHD (AU)	1.6 ± 1.0	1.6 ± 0.8	1.6 ± 1.3	1.7 ± 1.0	1.7 ± 0.7	ED vs. CD: <0.999ED vs. MF: 0.999ED vs. W: 0.997ED vs. ST: 0.993CD vs. MF: 0.996CD vs. W: 0.999CD vs. ST: 0.998MF vs. W: 0.968MF vs. ST: 0.965W vs. ST: <0.999	ED vs. CD: ≤0.001 trivialED vs. MF: ≤0.001 trivialED vs. W: -0.100 trivialED vs. ST: -0.110 trivialCD vs. MF: ≤0.001 trivialCD vs. W: ≤0.001 trivialCD vs. ST: -0.131 trivialMF vs. W: -0.084 trivialMF vs. ST: -0.085 trivialW vs. ST: -0.110 trivial
sHD (AU)	1510.0 ± 1517.5	1326.6 ± 1289.6	1534.0 ± 1496.3	1673.1 ± 1670.1	1309.2 ± 1233.1	ED vs. CD: 0.821ED vs. MF: 1.000ED vs. W: 0.865ED vs. ST: 0.862CD vs. MF: 0.687CD vs. W: 0.257CD vs. ST: <0.999MF vs. W: 0.895MF vs. ST: 0.772W vs. ST: 0.383	ED vs. CD: 0.130 trivialED vs. MF: -0.016 trivialED vs. W: -0.102 trivialED vs. ST: 0.141 trivialCD vs. MF: -0.146 trivialCD vs. W: -0.231 smallCD vs. ST: 0.014 trivialMF vs. W: -0.089 trivialMF vs. ST: 0.157 trivialW vs. ST: 0.237 small

wHD: weekly high intensity decelerations; mHD: monotony high intensity decelerations; sHD: strain high intensity decelerations; ED: external defender; CD: central defender; MF: midfielder; W: winger; ST: striker; AU: arbitrary units; ES: effect size

Greater wImp was found for ED than for CD, W, and ST (31%, 43%, and 50%, respectively). The mImp was greater for ED and CD than for ST (18% in both cases). Furthermore, the sImp was greater for ED than for CD, W, and ST (44%, 95%, and 91%, respectively).

No significant differences were found between positions for the different workload indices calculated for HA.

DISCUSSION

The aim of this study was to describe and compare the weekly acute load, monotony, and training strain of accelerometry-based measures across a professional soccer season (pre-season, first and second halves of the season) according to players’ positions. The main results revealed that among professional male soccer players, greater training monotony and strain (including impacts and HMLD) occurred in the pre-season than during the season. Moreover, such results were modulated by the player’s position.

Weekly high metabolic load distances varied between 5385 m and 7875 m. Similarly, in a study on 28 elite soccer players from a Dutch team intended to quantify training loads in relation to matches during a season, high metabolic load distances (between 565 m and 2472 m) were found from 4 days before the match day to the match day, reaching 6320 m in one week [27]. However, it is important to highlight that the specificity of exercise may constrain the load. In a study that analysed the metabolic distance in a single drill, variations between 357 m and 358 m were found [28]. However, this study was conducted in university students and might not reflect the actual load per exercise associated with elite soccer training. Biomechanical loads as high intensity accelerations and high intensity decelerations are of extreme importance in team sports due to the high demands of the “start and stop” actions that might harm athletes’ structures [29, 30].

Another study that collected data from 45 home matches over a three-season period found that the team completed a mean of 76 high intensity accelerations and 54 high intensity decelerations per match [31], which is in accordance with the present study, which found more weekly high intensity accelerations (~1000 n) than decelerations (~800 n). In contrast, a study conducted on 11 U-23 professional soccer players from the English Premier League found that players completed more high intensity decelerations (43 n) than high intensity accelerations (26 n) during a match [32]. This finding is in accordance with a systematic review of acceleration and deceleration profiles in elite team sports (including soccer) [30].

Regarding the number of impacts, the present study revealed a mean number of weekly team impacts of 1443.8. Between players, this value ranged between 1099.1 and 1617.6. Although little research has covered this topic [33], in the aforementioned study [32], it was found that during a match, the team suffered 6,172 impacts, which is higher than the values found in the present study if we consider the weekly acute impacts load.

Regarding overall accelerometry-based measures, our study revealed a simultaneous pattern between training monotony and training strain, by which it is assumed that when one increases or decreases, the other follows. This is in line with previous findings [15]. However, other work [17] has revealed contrary results, showing that higher training monotony is related to lower injury risk, while higher training strain values continued to be related to higher risk incidence.

The use of GPS devices to quantify elite athletes’ external load variations throughout a season (mainly associated with distance-based measures) are well documented in recent literature [19, 34, 35]. However, research remains scarce on accelerometry-based measures such as high metabolic load, high intensity accelerations and decelerations, and impact variations across a full-season period. Therefore, it is difficult to compare our results with those of other studies because the majority of the research regarding these metrics considers only the match demands and does not consider weekly variations [30], which is not the case in the present study.

Moreover, there is a tendency to analyse these metrics in a match [36] and compare data between match halves instead of comparing periods of a season [31, 32, 37]. Such studies have revealed no significant differences between the two halves of a match for high intensity accelerations or high intensity decelerations, though a slight decrease was found from the first half to the second half for both measures. In the present study, moderate differences were found between the pre-season and the first and second halves of the season for the number of impacts. Meanwhile, corroborating the trend observed in earlier studies of [31, 32], we found only small effect size differences for high intensity accelerations and decelerations between periods.

Considering variations in training monotony and strain, little evidence [22] exists supporting the type of calculations used in the present study for training monotony and strain (i.e., these calculations were applied through accelerometry-based measures instead of the well-established sRPE-based method) [18, 38]. Despite methodological differences, a previous study [19] found that training monotony tends to decrease across the season as the training strain seems to increase for the distance-based GPS device measures analysed. Although it is challenging to analyse training monotony and strain patterns across the season in the present study, it can be observed that more accentuated values tended to appear in the beginning and at the late phases of the season, while lower values were seen during the mid-season.

Regarding the differences between playing positions for acute load, training monotony, and training strain, only small effect size differences were found for all measures. Descriptive weekly acute loads of high metabolic load distances were greater for external defenders, midfielders, wingers, and strikers than for central defenders, while midfielders and wingers had significantly greater acute loads than strikers. Also, training strain values were greater for wingers than for central defenders and strikers, while no significant differences were found for training monotony. The acute loads of impacts were greater for external defenders than for central defenders, wingers, and strikers. Meanwhile, external and central defenders had greater training monotony values than strikers, while external defenders had greater training strain values than central defenders, wingers, and strikers. Also, central defenders had greater strain values than wingers and strikers, and midfielders had greater values than wingers and strikers. For high intensity accelerations, the external defenders had greater acute loads than midfielders. Furthermore, while no significant differences were found between positions in terms of training monotony, central defenders and wingers had greater strain values than strikers. For high intensity decelerations, external defenders had greater acute loads than central defenders, midfielders and strikers, and wingers had greater acute loads than central defenders and strikers. Also, while no significant differences were found between positions for training monotony, wingers had greater strain values than central defenders and strikers.

As for the abovementioned issues about studies considering only the match profiles of accelerometry-based measures, the same is true for differences between playing positions. However, in a study conducted on 46 professional soccer athletes, it was found that the central defenders had the lowest high metabolic load distances (1527 m) and the lowest number of high intensity accelerations (27 n) and decelerations (45 n) [39], which is in line with our results. However, the values found for high metabolic load for strikers contrast with the results of the present study, as they were significantly higher than what we reported.

Also, in a study comprised of 19 elite soccer players conducted over eight matches, it was found that the attackers had the lowest high intensity acceleration and high intensity deceleration values, while midfielders presented the highest values for both measures [40]. These findings are in line with our results, although it is difficult to compare them because the authors did not divide positions in the same way (i.e., they considered only defenders, midfielders, and attackers).

Also, as far as we know, no research has been done on weekly variations, between-period differences, and positional profiling for training monotony and training strain through accelerometry-based measures. For those reasons, in the present study, training monotony presented a between-week “w-shape” pattern across the full season for the overall accelerometry-based measures. Sudden large increases were observed for acute loads, training monotony, and training strain for the overall measures, especially in the late phases of in-season. Recent literature suggests that spikes of more than 10% in load do not necessarily cause injury [41]. Nevertheless, it seems to be preventive to control the load demands imposed on athletes and limit sudden increases or decreases in workloads between 10 and 15%, thus preventing injuries [42, 43].

The main evidence of the present study revealed that training monotony and strain were greater during the first weeks of the season and then dropped during the season and increasing again in the late phase of the season. Greater training monotony and strain were found in the pre-season than in the first and second halves of the season, but no significant differences were found for any other measures. Central defenders exhibited the lowest values of high metabolic load distances, while external defenders and midfielders presented the greatest numbers of impacts over the season.

The present study has some limitations, one of the most notable of which is the size of the sample. Also, we did not consider some potentially important metrics, such as player load, fatigue index, and dynamic stress. However, including too many variables could have hindered the analysis of this study. Despite these limitations, to the best of our knowledge, this was the first study to analyse variations of training monotony and training strain between periods of a full soccer season; it was also the first to observe variations between playing positions through accelerometry-based differences over an entire season. Future studies are needed to corroborate these findings, as it would be interesting to investigate the training monotony and strain differences between different playing formations.

CONCLUSIONS

The present study provided reference data of variation in external training loads for weekly periods within the annual season. Additionally, it was found that among professional male soccer players, greater training monotony and strain (including impacts and HMLD) occur during the pre-season than during the season, and these results were modulated by playing position. The current findings may help coaches and practitioners better regulate player load across the season in order to better prepare them for competition and reduce injury risk.