Analysis of professional soccer players in competitive match play based on submaximum intensity periods

Introduction

Knowledge of players’ activity during competitive match play is vitally important for subsequent prescription of specific training (Riboli, Esposito & Coratella, 2021). For this purpose, the use of global positioning systems (GPS) offers us valid and reliable information (Harper, Carling & Kiely, 2019; Beato et al., 2018) on distance and other derived variables (Gabbett, 2016), enabling us to study and monitor training and match play (Casamichanan et al., 2013; Gómez-Díaz et al., 2013).

Previous studies (Suarez-Arrones et al., 2015) have analysed the physical activity profile of professional soccer players in competitive match play, reporting the information as mean values (for example, metres per minute or metres sprinted per minute). However, owing to the intermittent nature of play in team sports (Johnston, Gabbett & Jenkins, 2014; Johnston et al., 2018), these mean values may underestimate the demands to which soccer players are subjected in certain phases of play (Whitehead et al., 2018), and if they are used as an intensity benchmark in designing training activities this could lead to insufficient preparation of players (Riboli, Esposito & Coratella, 2021). For this reason, in recent years there has been a great increase in research on maximum intensity periods (MIPs), which can be identified as the phases of play in which players show the highest level of conditional exertion. MIP analysis therefore offers us useful information on the maximum demands on athletes (Lord et al., 2020). The characteristics of these periods have been analysed in match play (Casamichana et al., 2019; Campos-Vázquez & Lapuente-Sagarra, 2018), and positional differences have also been evaluated (Delaney et al., 2018; Martín-García et al., 2018; Oliva-Lozano et al., 2020). In addition, recent research has compared the exertion of these periods in various training activities with that displayed by soccer players in competitive matches (Martín García et al., 2020; Martin Garcia et al., 2019). In these studies, a range of criterion variables, time windows (Delaney et al., 2018; Martín-García et al., 2018; Martín García et al., 2020; Martin Garcia et al., 2019) and methods of analysis (Whitehead et al., 2018) are used to calculate MIPs, always obtaining higher values when the time window used is smaller (Casamichana et al., 2019; Oliva-Lozano et al., 2020). Moreover, it seems clear that the demands on athletes are underestimated when segmental analyses are used instead of rolling average techniques (Campos-Vázquez & Lapuente-Sagarra, 2018; Cunningham et al., 2018) and that these MIPs seem to be context-dependent (Oliva-Lozano et al., 2020; Novak et al., 2021; Riboli et al., 2021).

One of the limitations that arise when calculating MIPs is that they refer to a single event (period) that occurs during play, and therefore do not provide information on time periods during the match when exertion was high, but not maximal. For this reason, it seems useful to undertake an analysis of the external load that soccer players accumulate in these sub-maximum intensity periods (SubMIPs), since they could be related to the levels of fatigue shown in matches (Schimpchen, Gopaladesikan & Meyer, 2020). Analyses of this kind have previously been performed with internal load variables, such as heart rate (Aşçı, 2016), or with other types of methodology that analyse the external load in match play in different ranges of intensity (Riboli, Esposito & Coratella, 2021). Moreover, the characteristics of SubMIPs have recently been studied in other team sports such as rugby, Australian rules football (Johnston et al., 2020) and futsal, both in competitive match play (Illa et al., 2020) and trainings (Illa et al., 2020).

The main objective of this study is to analyse SubMIPs manifested by professional soccer players during official matches (number of events and time spent in each event), according to the player position, match halve and match, and also to group the players by cluster grouping according to their SubMip values during the competition.

Methods

Results

The results are shown as mean plus/minus standard deviation (Table 2). The highest average values SubMip were found in the variables AccDens and Mmin, and the lowest values in Sprint.

Those who exceeded the threshold in the HSR variable on the most occasions and for the longest time were MF and WD; by contrast, CD showed the lowest values in number of events and time above the threshold in the HSR, Sprint and HMLD variables. WD had the highest number and duration values in Sprint and HMLD. OMF showed higher values than the other positions in number of events and time spend above the threshold in the AccDens variable, as opposed to FW, who had the lowest values in this variable. In the MetPow and Mmin variables WD were those with the lowest time and duration values. The position that showed the highest values in these two variables was MF. The average values for number of AccDens events and duration of Mmin proved to be the highest in the first half of the match, and the values for number of Sprint events and duration in HSR were the lowest during the second half (Table 2).

The statistical analysis showed significant differences between the individual variables in the number of events in HSR (H = 27.805, p = 0.01), AccDens (H = 51.733, p < 0.001), MetPow (H = 74.44, p < 0.001) and Mmin (H = 66.751, p < 0.001) and in the time the player spent above the threshold in the same variables (HSR: H = 26.7, p = 0.014; AccDens: H = 49.455, p < 0.001; MetPow: H = 68.868; p < 0.001; Mmin: H = 63.655, p < 0.001) (Table 3). Significant differences were also observed between the variables according to the playing position for number of events in MetPow (H = 50.55, p < 0.001) and Mmin (H = 44.099, p < 0.001) and the time above the threshold in these variables (MetPow: H = 47.888, p < 0.001; Mmin: H = 42.328; p < 0.001) (Table 3). Similarly, differences were found between halves in number of events for AccDens (H = 5.797, p = 0.016), MetPow (H = 7.402, p = 0.007), Mmin (H = 6.05, p = 0.014) and time above the threshold in these variables (AccDens: H = 8.611, p = 0.003; MetPow: H = 8.068, p = 0.005; Mmin: H = 8.602, p = 0.003). Only were found differences between matches for events and time of Mmin (H = 30.971, p = 0.006, H = 32.582, p = 0.003) and in time of MetPow (H = 26.513, p = 0.022) (Table 3).

The subsequent post-hoc analysis showed differences between OMF and FW in AccDens time, also shows significant differences in the MetPow and m/min variables in number of events and duration between all positions except between OMF and FW, between OMF and CD differences were found only in the duration m/min (Table 4).

Analysing the first and second halves by post-hoc analysis, we can see differences both in the number of events of AccDens, MetPow and Mmin (Z = 2.408, p = 0.008; Z = 2.721, p = 0.003; Z = 2.46, p = 0.007 respectively) and in time above the threshold in these same variables (Z = 2.934, p = 0.002; Z = 2.84, p = 0.002; Z = 2.933, p = 0.002 respectively), values for the first half being higher than for the second (Table 4).

In the clustering based on the time spent by the player in SubMIPs, three main groups were described (Fig. 1). In the first, the centroid was located in highest values in HSR, Sprint and HMLD. In the second, an accentuation of the AccDens, MetPow and Mmin variables was observed, and in the third and final cluster the centroid was located in lower values in each of the variables (Table 5).

When these clusters are compared to the positions used for the analysis, we can see behavioural tendencies during match play. In all positions except midfielders, we found a tendency to the third cluster (CD (68.4%), WD (74%), OMD (57.7%) and FW (64%)), MF tend to group 2 (50%). For CD, OMF and FW, the second option is group 2 (19.6%, 23.1% and 18.6% respectively), while for WD, group 1 (22%).

The ANOVAs revealed significant differences between the three clusters (p < 0.01), and the Levene F value indicated significant differences between the three clusters for each variable analysed (p < 0.01). The sampling adequacy measure was KMO = 0.64, for X² = 537.0 and p = 0.00 (Morissette & Chartier, 2013).

Discussion

The most important findings of this study are as follows: (I) the largest differences in physical demands based of SubMIPs in competitive matches are individual (HSR, AccDens, MetPow and Mmin); (II) there are differences between positions, between halves and between matches in the number of SubMIP events and the duration of these events in different variables; (III) the largest differences between positions are found in the MetPow and Mmin variables, while the differences between halves are most evident for AccDens, MetPow and Mmin; (IV) the physical demand is not determined solely by the position of a player, although this determines his way of acting during matches.

Recent research (Riboli, Esposito & Coratella, 2021) analysed the percentage distribution of intensity compared to MIP 1’, concluding that the average data in this distribution were much lower that the MIP 1’ values, especially for high-intensity exertion (9–19%). This type of analysis has also been considered in sports such as rugby and Australian rules football (Johnston et al., 2020), as well as in much more closely related sports such as futsal, both in match play and in training (Illa et al., 2020; Illa et al., 2020). The study cited (Illa et al., 2020) showed that high-intensity (80%) and very high-intensity (90%) exertions in competition were lower in high-speed actions (in this case only HSR) (0.17 ± 0 in both categories) and higher in distance variables (3.17 ± 2.32 and 0.75 ± 0.35 respectively) and acceleration variables (2 ± 1.12 and 0.67 ± 0.55 respectively). As these are different sports, these results cannot be compared to those of our study.

To the best of our knowledge, this is the first study to date that has analysed the number of events and the time spent in SubMIP in professional soccer matches. The most notable results at the individual level show significant differences (p < 0.01) between players in HSR, AccDens, MetPow and Mmin events, as well as in the duration of each. These results suggest that we should consider the physical demands of players in competitive match play individually. Also extrapolating this to training sessions, individualizing them, and considering SubMIP as one more tool for the control of the load.

The differences in physical demands between positions in match play have been extensively studied with classical variables analysis using absolute speed thresholds (Mallo et al., 2015; Di Salvo et al., 2007) and individual relative thresholds (Abbott, Brickley & Smeeton, 2018) or with analysis of MIPs in various time windows (Martín-García et al., 2018; Di Mascio & Bradley, 2013). Regardless of the method used, all these studies showed significant differences between positions. In our study, analysing of SubMIPs, we found significant differences in the AccDens variable (only in its duration), which was higher in OMF than in FW. The greatest differences between positions occur in the MetPow and Mmin variables in number of events (Table 3), except between OMF in comparison with FW and CD. These differences could be partially explained by position-dependent nature in soccer (Martín-García et al., 2018; Di Mascio & Bradley, 2013; Mallo et al., 2015; Di Salvo et al., 2007; Abbott, Brickley & Smeeton, 2018); indeed, using individualized threshold values for the SubMIP based on the MIP for each player, we can find significant differences between positions in the number of events and the time in SubMIP.

The results obtained show higher values in the first half than in the second for AccDens, MetPow and Mmin, in number of events (Table 3), with no significant differences between halves in the other variables analysed. Previous studies also showed differences between halves when performing an analysis using classical variables, a lowering of total distance (Di Salvo et al., 2007) and a reduction of the distance covered in various speed ranges in the second half (Rampinini et al., 2007), though not systematically. Moreover, when MIPs were analysed, differences were found in variables such as average metabolic power, and these differences were more evident when the time window analysed was larger (Casamichana et al., 2019). Despite the apparent reduction in physical performance in the second half shown by these studies, we should treat this information with caution, as the reduction in performance could be a result of the players’ accumulated fatigue, but it could also be due to the reduction in useful playing time (Rey et al., 2011) or other reasons.

On the other hand, there were almost no differences in of SubMIPs when a comparison was made between matches. Significant differences were only apparent in the duration of MetPow and in number of events and duration of Mmin. This is contrary to what is reported in previous studies that have analysed the same phenomenon from a traditional perspective, where greater differences are shown, especially in metres covered at high intensity (Oliva-Lozano et al., 2020). The individualization of the thresholds for analysis of the number of events and duration could explain this absence of differences between matches.

Clustering produced three major behavioural groups. The players of different positions are grouped in group 3, where the demands in all the variables are lower, suggesting that players who, for various reasons, have not undergone great conditional exertion during a match are located here, considering the number of SubMIP events during the competition, it seems normal. In the first group, the centroid was located in highest values in HSR, Sprint and HMLD, we could call it the sprinter group, WDs are placed as a second option.

Finally, group number 2, an accentuation of the AccDens, MetPow and Mmin variables was observed, and the MF are placed in it as the first option and the CD, OMF and FW as the second option.

An important factor to take into account when interpreting the results is the methodology used to calculate SubMIPs. In our study, the average of the three highest MIPs for each of the variables was individual, and therefore so were the thresholds for the SubMIP. This means that a player who obtains higher maximum peaks in a specific variable will have higher SubMIP thresholds than another player with lower MIP values, giving different possibilities in the relationship between MIP and SubMIP, depending on the variable and the player’s physical demands during the match. On this context, there are players with high MIP and many SubMIP events, for example, WD in HSR, or both low values like CD in Sprint as the nature of their position requires them to use or not certain demands. Players with lower MIP and higher SubMIP periods (such as CD in AccDens, probably because of the low SubMIP threshold) and players with high MIP and low SubMIP, are rarely able to exceed the SubMIP range, either due to the low frequency with which the actions are requested or due to the very high threshold at which these actions are found.

It is important to bear in mind that an athlete’s conditional demands in competitive play, and consequently the MIP attained, is multifactorial (Novak et al., 2021); this is why the average of the three highest MIPs for each player in each variable was taken, so as to minimize the possible noise that may be caused in the MIP by context (Illa et al., 2020). Given that players exceed the SubMIP theshold approximately 0.2 times per half in HSR, Sprint and HMLD, we can see that MIP analysis from a non-dynamic perspective, analysing matches independently, may underestimate the real maximum capacities of our athletes. An analysis over the course of different matches and training sessions may offer us a view closer to the real maximum demand of players, therefore, it could be useful to standardiHe the criteria for obtaining SubMIPs and explore this field in greater depth so as to obtain more detailed knowledge of player’s physical demands from this new perspective, with the aim of providing valid and reliable tools for designing training and analysing their conditional performance in competitive match play.

However, main limitations of this study were the sample size (a single professional team) and the number of matches analysed, so these results should be interpreted according to this specific competitive context.

Conclusions

In conclusion, the analysis of the SubMIP in professional soccer players during official matches shows individual-dependent differences in the variables HSR, AccDens, MetPow and Mmin. There are also differences between positions in MetPow and Mmin; AccDens, MetPow and Mmin for halves; and MetPow and Mmin for matches. Therefore, we can conclude that physical demands are not determined solely by a player’s position although their position may influence their playing profile during matches.

Practical Applications

Just as the use of MIP for monitoring load and analysing match play has brought about a change in training design, the use of SubMIPs to discover and quantify the physical demands of professional soccer players may be a useful tool that could help to optimiHe players’ performance.

Bearing in mind the individual differences in the characteristics of SubMIPs during match play in professional soccer, it seems appropriate to individualized training according to the profile of each player and to the relationship of MIP to SubMIP profiles, as this could be an indicator of the physical demand required and whether this leads to the desired adaptations in the athletes. In the same way, we can find different profiles of athletes, as we have been able to see in the groupings and in the differences of the possible MIP-SubMIP relationships. It seems interesting to also take this into account for the design of training sessions.

Supplemental Information