Passing behaviour patterns in UEFA Champions League finals (2018-2022) 

Adrián Torregrosa-Domínguez

Jesús Salado-Tarodo

José Flores-Rodríguez

Eduardo José Fernández-Ozcorta

Abstract

The aim of this study was to analyse the behavioural patterns of passes made in UEFA Champions League finals played between 2018 and 2022, and to identify the situational (field position, opponent pressure) and behavioural (passing technique, decision making) factors associated with successful passes. Successful passes are those that result in the loss of possession by the opponent, or culminate in a goal or a shot. The study was carried out on a one-off, nomothetic and multidimensional basis: it was based on the observation of a specific moment without continuous time tracking, the comparison of behaviours of seven teams, and the analysis of various levels of response with an observation instrument. To this end, an observation instrument was constructed and validated through expert review, pilot testing, and reliability and validity analysis, therefore ensuring accuracy in coding. The participants of the study were professional teams playing in the UEFA Champions League finals. In total, 4,658 passes were recorded and coded. Preliminary results indicate that passes from the offensive zone (third quarter into the opponent’s half) are more likely to lead to a goal or a shot, while passes from a greater distance are associated with a loss of possession. These findings suggest that timing, location on the field and passing distance are key factors in the success or failure of plays.

Introduction

Research into team sports has become increasingly important from physical, technical and tactical perspectives. This has enabled teams to make informed decisions in different situations, backed by scientific evidence (Rennie et al., 2018; Young et al., 2019). In this sense, it has become increasingly important in recent years to validate instruments that allow the assessment of competence in professional football, using data providers such as WyScout (Sánchez-López et al., 2023). Furthermore, possession transitions has been studied to better understand the probability of success in specific plays (Castellano-Paulis et al., 2009).

Observational methodology has been essential to this advancement, especially in the evaluation of behaviours in invasive team sports. Specifically in the field of football, studies have addressed technical-tactical actions, the evolution of goals in world competitions and ball circulation in various categories (Gréhaigne et al., 2010; Iván-Baragaño et al., 2022 Mićović et al., 2022; Muriarte Solana et al., 2023; Ortega-Toro, 2019). These approaches have provided valuable information on game development, effectiveness in offensive phases and optimisation of game strategies.

Observational methodology, in addition to being widely used, has paved the way for the implementation of various data analysis techniques in the study of team sports (Barreira et al., 2020). Among these techniques, polar coordinates (PC) have been fundamental, as they allow the estimation of relationships between a specific behaviour and other observed behaviours. Pedagogical proposals and the evaluation of performance indicators have been utilised in the analysis of offensive play (Flores-Rodríguez, 2020; Maneiro et al., 2018).

In the specific context of football, passing network analysis has been a valuable resource for defining team characteristics and explaining team success on the field (Buldú et al., 2019; Castañer et al., 2016; Maneiro et al., 2018; Zeng & Zhang, 2022). These studies have provided parameters such as the offensive behaviour index and the game control index, crucial for detecting the degree of a team’s control over the opponent.

In addition to the overall analysis, the importance of examining the effectiveness of possession units has been noted. These segments of play, defined by ball control, are crucial to analysing the quality, effectiveness and distribution of possession. Factors such as total number of passes, passing accuracy and other aspects have been shown to be closely related to the success of these units (Collet, 2012; Hewitt et al., 2016; Zeng & Zhang, 2022). These studies have provided valuable information on the relationship between different parameters and success in the game.

In relation to passing and scoring, the relationship between the number of passes and scoring success has been found to present some contradictions. For example, while it is suggested that fewer passes per action increases the probability of scoring a goal, it has been observed that 80% of possession units that end in a goal involve more than three passes (Aguado-Méndez et al., 2020; Alves et al., 2023; Taha & Ali, 2023). These discrepancies emphasise the importance of defining the type of attack and the role of counter-attacking actions in the context of the game (Chmura et al., 2021).

The incidence of passing errors is a key element to consider, given that the majority of possession units do not result in goals. It has been observed that short passes can reduce losses and thus improve the chances of successful play (Chmura et al., 2021).

On the other hand, analysis of ball recovery zones has revealed that recovering the ball close to the opponent’s goal is associated with an increase in the probability of scoring. Furthermore, counter-attacks generated in the central channel have shown a correlation with successful play (Mendes & Morante, 2011). These findings underline the relevance of recovery strategies and their impact on the development of play.

Finally, it is essential to consider the influence of situational variables on the performance and behaviour of players. Elements such as the location of the match, the level of the opponent and the match status have been shown to have both physical and tactical effects (Mackenzie & Cushion, 2013; Taylor et al., 2008). These variables have been shown to influence aspects such as time of possession, type of pass and possession success rate, and have varied significantly between the first and second halves of matches (Maneiro et al., 2021).

Based on the current state of the evidence, the aim of this research was to analyse the passing behaviour patterns in the UEFA Champions League finals played between 2018 and 2022. Specifically, passes that ended in a shot and passes that ended in loss of ball possession were studied according to time of play, score, type of pass made and zones of the field where the passes were made and received. 

Method 

Materials

For this research, an ad hoc observation instrument was designed to record relevant behaviours in relation to the objective of the study. The construction and validation of the instrument was carried out in three main phases: initial design based on the literature, pilot testing and adjustment through expert judgement. The observer training process and the reliability and validity analyses were prescriptive to ensure the accuracy and utility of the instrument.

Observational design

The research was conducted using a one-off, nomothetic and multidimensional observational design (Anguera & Hernández-Mendo, 2013). One-off, because the matches observed correspond to specific moments without continuous time tracking; nomothetic, because different units of analysis were compared—in this case, the behaviours of seven different teams; and multidimensional, because several levels of response, collected through the observation instrument, were studied.

Participants 

Data were collected from five different UEFA Champions League finals (Table 1), all played on neutral ground, between 2018 and 2022. In total, seven different teams from three different European leagues were analysed.

Table 1

Finals analysed.

See Table

In accordance with the Belmont Report (1978), it was not necessary to obtain informed consent or review by the relevant ethics committee because: (a) the study involved the observation of individuals in a public setting (football stadium); (b) the teams observed had no expectation of privacy as the matches were broadcast worldwide; and (c) the study did not involve direct intervention or interaction of the researchers with the athletes studied. 

The units of observation were all passes made by the different teams with the exception of those made by resuming the game (for example, goal kicks, free kicks, fouls, corners, throw-ins, etc.).  

Instruments 

An ad hoc observation instrument was designed to record relevant behaviours in relation to the research objective. The construction of the observation instrument consisted of three phases, based on the work of Aguado-Méndez et al. (2020). 

First phase. Two doctors in Physical Activity and Sport Sciences, with previous experience in observational studies, designed an initial version based on the available literature. In this first stage, we opted for a combination of the field format of Aguado-Méndez et al. (2020)with a system of categories. The playing field format was divided into a grid of 24 rectangular zones, organised into four rows and six columns, labelled with the prefix “Zone” followed by a number from 1 to 24. The field grid numbering started in the upper left corner of the field with “Zone 1” and proceeded from left to right and from top to bottom, ending in the lower right corner with “Zone 24”. Zones 1 to 12 were configured as the home camp, while zones 13 to 24 belonged to the opposing camp. This combination made it possible to take advantage of both the flexibility of the field format and the theoretical consistency of the category system (Anguera & Hernández-Mendo, 2013). 

Second phase. The instrument was subjected to a precautionary test (Anguera, 2003), consisting of the recording of several matches not included in the sample. The precautionary test served to modify the initial design of the research instrument by adding and deleting different criteria and categories. The test was considered as finished when, during the recording of the non-sampled matches, no behaviour was detected that could not be recorded with the research instrument. 

Third phase. The instrument was assessed by three experts, doctors and university lecturers in Physical Activity and Sport Sciences, who marked their agreement or disagreement with each of the categories and criteria of the instrument. Ultimately, all the criteria and categories that made up the final version of the instrument obtained a percentage of agreement of over 80%. The final instrument used for the observation is indicated in Table 2.

Table 2

Observation instrument.

See Table

After the tool design was finalised, it was implemented in Microsoft Excel for recording and coding the actions; it therefore functioned as a recording instrument. Polar coordinate analysis was applied with the HOISAN 1.2 software (Hernández-Mendo et al., 2012). Prior to the calculation of polar coordinates and as a requirement, sequential delay analysis was performed using the GSEQ 5.1 software (Bakeman & Quera, 2011). Finally, after polar coordinate analysis, the significant associations were represented with the program Snowflake.

Procedure 

The nature of the research, based on observations at football matches and the analysis of existing data, did not involve manipulation of participants or direct intervention in their physical or emotional integrity, thus avoiding the need for a bioethics committee for approval.

The recording of the actions was carried out by an observer, who participated in the design of the observation instrument.To optimise the reliability of the recordings, the observer participated in a training process, which consisted of recording matches not included in the sample. The training process was concluded when values equal to or greater than .8 were obtained for Cohen’s Kappa statistic at the intra-observer level, a near-perfect result (Landis & Koch, 1977). Once the training process was completed, the matches that made up the study sample were recorded.

Lastly, the results of these analyses were reviewed and the instrument was adjusted accordingly to ensure its accuracy and usefulness in measuring the desired constructs. This systematic approach ensures that the observation instrument is both reliable and valid for use in future research.

Data analysis 

The observational data were analysed using the polar coordinate technique, which allows the graphical representation of the activation or inhibition relationships between the analysed behaviours. This technique has been used in the study of different team sports, such as football (Castañer et al., 2016) or handball (Flores-Rodríguez & Alvite-de-Pablo, 2023). In this analysis, one of the behaviours assumes the role of focal behaviour, since it is considered to be the generator of the relationships with the rest of the behaviours that participate in the analysis, which assume the role of conditioned behaviours. 

As a prerequisite, it is necessary to perform the sequential analysis of positive lags, which will inform about the prospective perspective, and of negative lags, to learn about the retrospective perspective (Sackett, 1980). Once the sequential analysis has been performed, the Zsum statistic performs the integration of the two, and values are obtained that can have a positive or negative sign. The results were represented in one of four possible quadrants, depending on the combination of signs obtained in each Zsum. 

As established in previous research (e.g., Anguera et al., 2011; Camerino et al., 2019), the graphical combination allows us to explain how to interpret the associations between the focal behaviour, located in the centre of the figure, and the conditioning behaviours in each quadrant. The association is shown both quantitatively (vector length) and qualitatively in quadrants I, II, III or IV. If the relationship is located in quadrant I, it indicates a mutual activation relationship between the focal behaviour and the conditioned behaviour. Conversely, when the representation is in quadrant III, it indicates the existence of a mutually inhibiting relationship between the focal behaviour and the conditioned behaviour. The representation in quadrant II indicates that the conditioned behaviour triggers the occurrence of the focal behaviour while being inhibited by it. Lastly, placement in quadrant IV indicates that the focal behaviour inhibits the conditioned behaviour while being activated by it.

By presenting the analysis described above, the aim was to analyse the behavioural patterns of passes made in UEFA Champions League finals played between 2018 and 2022, identifying the situational (field position, opponent pressure) and behavioural (passing technique, decision making) factors associated with successful passes.

Results 

Significant associations are represented below: those with a radius greater than 1.96 (p < .05), identified between the focal behaviour and the conditioned behaviours located in quadrants I and III. Placement in quadrant I indicates a mutually activating relationship, while representation in quadrant III expresses mutual inhibition. To facilitate the understanding of the results, they are presented in two subsections: in the first one, the passes that preceded a shot as focal behaviour; and in the second one, the role of focal behaviour is assumed by the passes that ended in a loss of ball possession. 

Passes preceding a shot

To understand the behavioural patterns related to passes that preceded a shot, the combination of the categories GAS (passes made just before a shot that ended in a goal) and TMS (passes made just before a shot that did not end in a goal) was used as the focal behaviour. In Figure 1 (Graphs A, B and C), the categories belonging to the criteria of minute (MIN), score (SCO), and type of pass (TPM) assumed the role of conditioned behaviours in Graph A. On the other hand, in Graph B the conditioned behaviours were the categories of the criterion of pass initiation zone (PIZ); and in Graph C, the categories of the criterion of pass reception zone (PRZ) were considered as conditioned behaviours. 

With regard to the minute criterion (MIN), it can be observed that the passes that preceded the shots presented a mutual activation relationship with the period T06, the period of play between the 75th and 90th minute, and with the period ET2, extra time in the second half. On the other hand, there is a relationship of mutual inhibition with the passes made between minutes zero and 15 (T01) and with those made between minutes 31 and 45 (T03). In relation to the behaviours belonging to the criterion of match score (SCO), a relationship of mutual activation can be seen with the LOS behaviour (the team that makes the pass is losing), and a relationship of mutual inhibition with the DRA behaviour (the team that makes the pass is drawing). Finally, no significant relationships were found between passes that preceded a shot and behaviours corresponding to the type of pass made (TPM). 

In turn, the significant associations found between passes that preceded a shot and behaviours related to the pass initiation zone (PIZ) are depicted. A mutual activation relationship was found with field zones I14, I18, I19 and I22, and a mutual inhibition relationship with the following zones: I03, I06, I07, I08 and I10. 

Significant relationships between passes that preceded a shot and the pass reception zone (PRZ) have also been reflected. Mutual activation relationships with zones R18, R19, R20, R22, R23 and R24 stand out. In contrast, mutual inhibition was found within zones R02, R03, R06 and R07.

Passing that resulted in loss of possession

The LBT category was used as the focal behaviour to identify behavioural patterns related to passes that ended in a loss of ball possession. In Figure 2 (Graphs D, E and F), the conditioned behaviours in Graph D were those belonging to the criteria of minute (MIN), score (SCO), and type of pass (TPM). In Graph E, the categories of the criterion of pass initiation zone (PIZ) assumed the role of conditioned behaviours. In Graph F, those categories corresponding to the criterion of pass reception zone (PRZ) were considered as conditioned behaviours. 

In relation to the time criterion, it can be seen that the passes that preceded shots presented a mutual activation relationship with the ET2 period (extra time in the second half). A mutual inhibition relationship can also be observed with passes made between the 45th and 50th minute (T04) and those made between the 60th and 75th minute (T05). With respect to the behaviours belonging to the criterion of score (SCO), the results indicate a mutual activation relationship with the LOS behaviour (the passing team is losing), and a mutual inhibition relationship with the DRA behaviour (the passing team is drawing). Lastly, regarding the type of pass made (TPM), mutual activation was found with MTP (the pass crosses only one zone) and LTP (the pass crosses more than one zone), and a mutual inhibition relationship with STP (the pass does not cross any zone).

Similarly, the significant associations found between the focal and the conditioned behaviours belonging to the criterion PIZ (the pass initiation zone), are shown. Mutual activation was found with I01, I04, I05, I07, I08, I24, and mutual inhibition with the following: I13, I14 and I15.

In turn, the relationships found with the behaviours corresponding to the criterion PRZ (the pass reception zone), are represented. In relation to the following categories, a relationship of mutual activation was observed: R01, R08, R12, R22 and R23.

Discussion 

The aim of this research was to analyse the passing behaviour patterns in the UEFA Champions League finals played between 2018 and 2022. Specifically, passes that ended in shots and passes that ended in loss of possession were studied according to time of play, score, type of pass made, and the zones of the field where the passes were made and received. After analysis of the data collected through the polar coordinate technique, each of the variables was arranged in one of the four quadrants of the polar coordinate map, which allowed us to describe the relationship of the variable with the focal behaviour. 

As for the study of passes ending in shots, the focal behaviours TMS and GAS were analysed by collecting both behaviours where a shot occurred. In this respect, in terms of the “Minute” dimension, our results show a relationship of activation of these behaviours with ET2 and T06, marking a tendency towards the appearance of shots in the last minutes of the match. This is probably related to periods in the match when the players are more physically and mentally exhausted. In contrast, in the first minutes of the match, when exhaustion should be less, there is a relationship of mutual inhibition as reflected in the data of T01 and T03. The “Result” dimension reflects a mutual activation relationship between TMS/GAS and LOS, as well as inhibition with the variable DRA, which is contrary to the findings of Maneiro et al. (2021), where the highest percentages of successful possessions were related to teams that were winning or drawing. Regarding the dimension of “Pass initiation zone”, there are mutual activation relationships with variables I14, I18, I19 and I22, areas of the field where the aim is to create spaces between lines to create advantageous situations in which passes can be made to areas closer to the goal for shots to take place. These results are in line with those reported by Immler et al. (2021) in terms of the involvement of midfielders with successful possessions, who tend to participate within the reflected zones, as well as that described by Maneiro et al. (2020) on passes made in the last 30 metres by the winning teams. However, mutual inhibition relationships are established with the variables I02, I03, I06, I07 and I16, zones related to the initiation of play, which is consistent with the findings of Chmura et al. (2021) on the involvement of defenders in unsuccessful possessions, as they are zones related to the positioning of defenders. The dimension of “Pass reception zone” indicates mutual activation relationships with the variables R18, R19, R20, R22, R23 and R24; zones in the opposite field related to shots. In contrast, inhibition relationships occur with the variables R02, R03, R06, R07 and R16, areas of play initiation, as we have noted above.

As for the EPB focal behaviour, the analysis of the “Minute” dimension highlights a mutual inhibition relationship with T04 and T05, periods in the second half where teams are likely to take fewer risks in their passing as they seek to avoid losing situations that allow the opponent to create scoring opportunities. On the contrary, a mutual activation relationship is established with ET2, the final minutes where teams risk more as they seek to generate scoring opportunities, as previously reflected in the TMS and GAS analysis. The dimension “Result” points to a mutual activation relationship with LOS, reflecting the higher risk taken by losing teams. In contrast, a mutual inhibition relationship is generated with the variable DRA, probably for similar reasons to those expressed above in the results of T04 and T05. The dimension “Pass initiation zone” shows mutual activation relationships with zones I01, I04, I05, I07, I08 and I24, which correspond to the initial phase of the game and the central channel. In this area, ball losses are particularly dangerous, as they can generate counterattacks by the opposing team (Gómez et al., 2012; Mendes & Morante, 2011). Moreover, these zones are associated with the defenders, who, according to Chmura et al. (2021), are involved in unsuccessful possessions. The I24 zone is an advanced position on the field where there is usually a low density of attacking players, which makes it difficult to pass to close teammates. On the other hand, mutual inhibition relationships are observed with zones I13, I14 and I15, which are usually occupied by midfielders. These players, according to Immler et al. (2021), are frequently involved in successful possessions. The analysis of the pass reception zone variable produces mutual activation relationships of the EPB focal behaviour with R01, R08, R12, R22 and R23. These variables point to zones in the home half of the field related to the initiation of play and where error can lead to counter-attacking situations, as described above, and forward areas of the field, which in many cases involve more risky passes that make mistakes more common. In contrast, inhibition relationships occur with R11, R13, R14 and R15, which, as mentioned above, are areas related to the construction of the play and which involve the midfielders more, which Immler et al. (2021) relate to successes in possession. 

Finally, the analysis of the dimension “Type of pass” reflects mutual activation with MTP and LTP, passes in which the distances are greater, which increases the probability of error and interception by the opponent. On the other hand, there is a mutual inhibition relationship with STP, reflecting how shorter distance passes allow for greater accuracy and safety, as pointed out by Chmura et al. (2021), and relates to those teams that are winning (Praça et al., 2019).

Practical recommendations

The observation instrument that was developed proved useful in analysing passing performance and the influence of the match situation on passing decisions. The findings showed an increase in passes leading to goals in the final minutes of the match and a higher incidence of ball losses at these times, suggesting that teams take more risks in search of scoring opportunities. The importance of passing in the final third of the field, close to the opponent’s goal, was emphasised to generate successful shots. Although passing distance is not directly related to creating shooting situations, short passes reduce ball losses, suggesting a more conservative and accurate strategy.

With these results, specific training sessions can be designed to improve technical skills and prepare players for various game situations. These training sessions could focus on three main factors: the match situation, the passing trajectories and the areas of the field where passes are initiated and received. Integrating these elements into task design can be beneficial. For example, game situations of reduced play in specific zones of the field could provide opportunities for these three conditioning factors to be worked on. In these situations, exercises in a reduced space can improve the precision of short passes, even in situations with high player density. Moreover, if these are set up with different spatial dimensions and in different zones of the field, they are likely to improve the ability to defend and attack effectively in those areas of play. Lastly, incorporating hypothetical scenarios in which the team is drawing or losing allows for practice under pressure and the development of quick and effective decision-making. Here, mental preparation is crucial to handle pressure and keep calm in critical moments, with the use of short passes being particularly beneficial, as they provide greater security. 

Overall, the integration of these specific approaches could not only improve players’ technical skills, but also develop their ability to cope with tactical and emotional challenges during a match, which may contribute to stronger and more cohesive performance on the field.

Conclusion

This study presents significant contributions to the analysis of performance in football, although it faces some limitations. The main one was the lack of access to videos of the finals and full data packages of the passes, which limited the depth of the analysis and the validation of the findings. Future research with access to these resources could provide a more comprehensive and accurate analysis.

Despite these limitations, the observational methodology used proved to be an extremely useful tool in the context of scientifically analysed football. This methodology allows for a detailed assessment of player behaviour in real game situations. In particular, polar coordinate analysis was effective in identifying patterns of behaviour and relationships between variables, and provided a comprehensive view of performance in the field.

These findings can be of great use to top-level teams, especially in preparation for the final rounds of elite competitions such as the UEFA Champions League. The implementation of these evidence-based strategies allows for the development of more effective game plans tailored to the specific circumstances of the match. Understanding passing patterns and their relationship to successful play will allow for better tactical preparation, optimisation of ball possession and increased goal scoring opportunities.

Therefore, we believe that this work contributes to the field of performance analysis in football by providing a deeper understanding of how passing patterns influence game outcomes, which will help to devise more effective tactics. The validation of the observation instrument and the application of polar coordinate analysis offer robust analytical tools for future studies and sport professionals. In addition, the findings can influence training methodologies, improving passing accuracy and decision making, especially at critical moments of the game.

Finally, coaches can use these findings to improve the performance of their teams, making strategic and data-driven use of observations and analyses. This observational methodology provides a sound basis for informed decision-making and continuous improvement of team performance, and offers a significant competitive advantage in the elite football field.