Contextual Variables and Time-Motion Analysis in Soccer

match status, match location, opponent level match half effective playing time effective playing time home away losing drawing winning higher medium bottom 2ndhalf 1st half


Introduction ▼
Analysis of movement patterns during matchplay has been used to evaluate physical demands in soccer for more than 30 years [33] . However, the work-rate profi le can be altered as a result of many factors, including the method used by different systems of time motion analysis [29] . In this regard, a recent review suggests that both diff erent techniques used to analyse work rate and ' situational ' variables (such as match location or status and the quality of the opposition, among others) have an important infl uence on measures of soccer performance [6,17,39] . As such, the eff ective assessment of players or teams requires knowledge of the contextual factors [20,21,28,39] that can potentially aff ect performance [31] . One of these variables is the halves of the game . Previous studies [2,27,34] have reported that a longer distance is covered during the fi rst half of the match compared to the second half, although the results are not entirely consistent [4,11,41] . These inconsistent fi ndings are probably to do with the interaction of other variables such as respect to 4 contextual variables: match status, match location, opponent level and match half , which were analysed in relation to the eff ective playing time . A descriptive analysis and a multivariate mixed model were employed for the analysis of change processes in soccer. The distance total covered (m) by players at diff erent work intensities during the eff ective playing time was greater when playing at home (3 931 vs. 3 887 away ), when the reference team was losing players ' physical activities are due to the tactics and style of play adopted, although this variable must be studied independently. In the context of the above, the present research takes a novel approach in 2 respects. Firstly, it uses multivariate analysis to study the interaction between distances covered at diff erent work intensities and contextual variables. Secondly, it makes use of the eff ective playing time , i. e., the time when the ball is in play, in order to study the workload of professional soccer players. On the one hand, it is likely that the variability of results reported for the physical workload [20,22,27] of players in diff erent matches is due to variability in the total playing time . However, previous methods have not addressed the eff ective playing time , even though this tends to be longer during the fi rst half; this is in contrast to the total playing time, which is usually greater in the second half [9] . The distinction between eff ective and total playing time would therefore seem to be relevant in terms of elucidating the diff erential eff ects on players ' physical profi le. In summary, the aim of the study was to determine physical performance profi les in soccer by observing a professional team in competitive matches, analysed from the perspective of eff ective playing time (E t ). Specifi cally, the study sought to identify the interactive eff ects of match location (home vs. away), score (win, draw and lose), opponent level (top, medium and bottom) and match half (fi rst vs. second) on work-rates.

Material and Methods ▼ Participants
A multi-camera computerised tracking system ( AMISCO Pro ® , version 1.0.2., Nice, France) was used to gather data from Spanish Premier League players completing an entire match half (434) during the 2005 -06 season. Data were collected for the duration of each match half, including injury time. The reliability and validity of this semi-automatic tracking system has been evaluated in other studies [11,12,41] . The club in question gave permission for this information to be used. However, to ensure team and player confi dentiality, all data were desensitised before analysis and were processed in accordance with the Declaration of Helsinki [15] . Institutional approval for the study was given by the ethics committee of the University of the Basque Country.

Total and eff ective playing time
Total playing time (T t ) was defi ned as the duration of the match as a whole, including stoppage time. Eff ective playing time (E t ) refers to the duration of play after subtracting the time taken up by stoppages, substitutions, goals and injuries, etc., in other words, subtracting the amount of time in which the ball is out of play.

Contextual variables
4 independent variables were included in the research. With respect to the contextual variable match location , and in line with previous studies [2] , we distinguished between matches played at home and away. As regards the match halves these were also divided into 2 levels: fi rst and second half. With respect to the opponent level we examined diff erences in physical performance when the reference team played against successful teams (ranked in the top 6 league positions), moderately successful teams (ranked 7 th to 13 th in the league) and the least successful teams (ranked in the bottom 7 of the league). These categories are similar to those reported previously [28,39] . With respect to the partial and fi nal result or match status [21] , this was divided into 3 levels, i. e., whether a team wins, loses or draws in each half.

Statistical analyses
Data are presented as the mean ± standard deviation ( ± SD ), along with the 95 % confi dence interval (95 % CI). Initial statistical analyses were performed using SPSS for Windows version 17.0 ( SPSS Inc., Chicago, IL, USA). Diff erences in match time ( total and eff ective ) with respect to the 4 independent variables were determined using the Student ' s t -test ( halves and match location ) and a one-way analysis of variance ( opponent level and score ). When a signifi cant F-value was found, Bonferroni ' s posthoc tests were applied. The level of statistical signifi cance was set at p < 0.05. A multivariate mixed model using SAS for Windows 9.1 [35] was also applied to analyse the change processes in soccer. A linear mixed model is a parametric linear model for clustered, longitudinal or repeated-measures data that quantifi es the relationships between a continuous dependent variable and various predictor variables. Longitudinal data diff ers from traditional multivariate data, in which a number of measurements are collected for each subject before analysing the multiple measurements as a single multivariate outcome. Here, Mauchly ' s test of sphericity was used to determine, fi rstly, that the data could be treated as multivariate (transformed variables, χ 2 = 21 675.6; Pr > ChiSq < 0.0001) and, secondly, that they fulfi lled the criterion of orthogonality ( χ 2 = 8 340.2; Pr > ChiSq < 0.0001). Both tests were signifi cant, thus confi rming the suitability of a multivariate procedure.

▼ Total and eff ective playing time
The average total duration (T t ) of each half was 46 min 32 s ( ± 44 s) for fi rst halves and 48 min 35 s ( ± 1 min 4 s) for second halves, while the mean eff ective playing time (E t ) in each half was 26 min 19 s ( ± 2 min 39 s) and 26 min 4 s ( ± 2 min 25 s), respectively ( • ▶ Table 1 ). The time during which the ball was in play did not reach 55 % ( E t / T t % ) of the total match duration, independently of the contextual variables. There were signifi cant diff erences in the T t with respect to the variables halves, match location and score , as well as in the E t with respect to the variable opponent level .

Distance covered
When considering the whole match, the total distance covered by players in each half ranged between 3 871 m and 3 958 m during the eff ective playing time, which corresponds to 69 % of the distance covered in the total playing time (5 667 ± 450 m). The mean total distance ( ± SD ) covered under diff erent running intensities for each half is shown in • ▶ Table 2 . The time motion analyses revealed that during 45 min the players performed 117 m of sprinting (SpD) and 145 m of very high-intensity activity (VHD). In the E t the ' relative weight ' ( % ) of the distance covered at 5 of the 6 intensities (SpD, VHD, HD, MD and LD) was greater than the corresponding fi gure for the T t . The opposite occurred for the variable SD, whose value was lower in the E t .

Distances covered and contextual variables
The initial descriptive analysis ( • ▶ Table 3 ) shows the mean distances covered in metres (m), standard deviations ( ± SD ) and confi dence intervals (95 % CI) according to the diff erent movement categories and for the eff ective playing time (E t ) with respect to the situational variables. Mauchly ' s test was used to determine whether or not the withinsubjects variance-covariance matrix had a Type H covariance structure [16] . The results for the type III general linear model (GLM) procedure for within-subjects eff ects showed a similar trend. Except for match location the remaining contextual variables were shown to be signifi cant when analysing workload during the E t ( • ▶ Table 4 ). This indicates that players of the observed team cover diff erent distances at diff erent intensities depending on the situational variables. As regards the multivariate model used, the results of the type III (fi xed eff ects) analysis showed signifi cant diff erences between the distances covered at diff erent intensities during the E t for the variables halves, opponent level and score (see • ▶ Fig. 1 ).

Discussion ▼
The purpose of this study was to determine any qualitative and quantitative diff erences in the demands placed on elite soccer players during match-play according to eff ective playing time  and with respect to 4 situational variables ( halves, score, opponent level and match location ) that were analysed simultaneously.
Our results are consistent with recent investigations using sophisticated measurement technologies [10,20] and which demonstrate that the mean distance covered by male elite outfi eld players for each half is ~ 5 600 m in T t , but only ~ 3 900 m in E t ( • ▶ Table 2 ), equivalent to 69 % . In the T t , our data show a workrate profi le similar to that reported in other studies of Spanish Premier League players [10,12,20,41] . Mean values for physical demands are also close to the mean values obtained in studies of players in the Italian Serie A [10,22] , the English Premier League [4,10] , the German Championship [10] and the Swedish professional league [1] . However, they contrast with those reported by Rienzi et al. [34] , who found that international South American players covered less total distance during a game. The present paper is the fi rst to report external workload distances for intense movement when taking into account the eff ective playing time (E t ). The E t accounts for a little over 50 % of the total match time (T t ), it being the only time during which teams have the opportunity to alter the score. In support of this notion a time-motion analysis based on E t ( ~ 70 % of workload corresponds to this period) can provide more precise information about a player ' s physical activity, which may have direct repercussions on the match outcome. In this regard we found significant diff erences with respect to the duration of the T t but not for the E t . This suggests that it would be useful to evaluate players ' activity on the basis of the E t , since this measure remains more stable across matches ( • ▶ Table 1 ); this would eliminate the possible variability between matches that is associated with the T t , such as in matches where the home team is losing and the referees prolong the T t [36] . The second noteworthy aspect of the present study is that we applied a multivariate analysis including contextual variables, those hypothesised to aff ect the physical performance of players. The players ' work-rates showed some signifi cant diff erences in relation to the situational variables ( • ▶ Table 4 , within-subjects eff ects ). In soccer, the evidence for a diff erence in the total dis-   tance covered between halves is inconsistent, and a signifi cant decrement does not necessarily occur in all players [6] . Some studies [2,5,11,22,34,30] have reported that the distance run decreases during the second half, suggesting a form of fatigue [32] . Indeed, recent studies have shown that the amount of both high-intensity running and sprinting declines as a soccer match progresses [14,18] . However, Di Salvo et al. and Zubillaga [11,41] reported just the opposite, while Bradley et al. [4] found no differences for high-intensity running and sprinting. In the present study, signifi cant diff erences were found between the fi rst-and second-half movement patterns for players ' workload in the E t . These diff erences in workload occur despite the fact that the duration of the E t is similar in the 2 match halves ( • ▶ Table 1 ).
With respect to match location (home vs. away), no signifi cant diff erences were found for distances covered at diff erent intensities. Contrary to the fi ndings of Lago et al. [20] the concept of home advantage does not seem to have repercussions for physical performance ( p = 0.06). The absence of any diff erences could be due to the infl uence of the interaction with the other contextual variables. It should also be noted that for match location , no signifi cant diff erences were found in the duration of the E t between the games analysed ( • ▶ Table 1 ). With respect to the opponent level , and in line with the fi ndings of other studies [22,28] , the poorer the quality of the opponent, the shorter the distance covered by the reference team. However, in contrast to Lago et al. [20] , who found that teams playing against better quality teams ran less distance at low intensities (0 -11 and 11.1 -14 km · h − 1 ), the present results show that when playing against more successful teams the reference team covered greater distances in all intensity range categories (Top > Medium > Bottom) except for the SpD category (Medium > Top > Bottom), although signifi cant diff erences were found for the E t ( • ▶ Table 1 ).
Finally, the physical profi le was also infl uenced by match status . Contrary to Zubillaga [41] but similar to Lago et al. [19,20] it seems that the distances covered (Spd, VHD, HD and MD) by the reference team were greater when the result was adverse ( • ▶ Table 4 ). This suggests that when losing, players try to reach their maximal physical capacity in order to draw or win the match. Accordingly, players clearly performed less low-intensity activity. Furthermore, this occurred when no signifi cant diff erences were found for the E t ( • ▶ Table 1 ).
A limitation of the present study is that the players ' position was not taken into account, in contrast to the approach taken in some previous reports [3,4,12,13] . It is likely that the workload of players according to their position is aff ected diff erently depending on the contextual variables involved. It has already been noted that research fi ndings and conclusions often vary across independent studies. Certainly, no one study can measure and control for all extraneous infl uences, particularly when results may be infl uenced by diff erent contextual variables [19] that aff ect players ' performance, for example, the type of competition [41] , the players ' level [22] , the playing style of diff erent leagues [34] , the match status or the quality of the opponent [20] . The present study only provides a simple overview of the work-rate profi les of elite soccer players by analysing longitudinal data, although it is the fi rst to report a MANOVA analysis. The particular applications of this analysis are useful for identifying contextual dynamics, and it off ers empirical clues to the infl uence of multivariate factors that should not be considered in isolation. If performance is adversely infl uenced by specifi c situational variables, possible causes can be examined and match preparation focused on reducing such eff ects [20] . Identifying physical qualities is a sine qua non among the other attributes needed by athletes to be competitive in the teamsports arena. The present fi ndings suggest that a failure to consider the E t when quantifying the physical workload of players may also aff ect this. Furthermore, consideration of the eff ective playing time provides more precise information about competitive physical demands, and this can then be applied to the training context in order to develop drills, etc. that are more closely tailored to actual match requirements [8] . Our results highlight a number of variables that could explain physical workload in soccer players, and combinations of these variables could be used to develop a model for predicting (from a probabilistic viewpoint) the physical activity profi le in competition. Indeed, the fi ndings of this study, together with those of other authors [19,21,39] , suggest that eff ective assessment of soccer performance at a behavioural level needs to account for the potential interactions between situational variables. It is hoped that the present fi ndings will serve to broaden the body of research on physical demands in elite soccer match-play, as well as improving knowledge of specifi c situational variables and their possible infl uence as regards tactical preparation for matches. As such, the results could be used to reduce undesirable eff ects [7] (for example, by preventing a decline in players ' performance or avoiding an increased likelihood of injury) or to develop recovery strategies that help players to maintain their performance in soccer.