Development and initial validation of a defensive pressure index using tracking data in the Chinese super league

Abstract

Tracking data have enhanced the analysis of football performance; however, defensive pressure remains insufficiently operationalised. This study aimed to develop a Defensive Pressure Index (DPI) based on movement tracking data, to examine its association with defensive outcomes, and to explore whether DPI differs across running score. Positional tracking data from 30 Chinese Super League matches during the 2018 season were analysed, yielding 2,032 defensive sequences. Defensive pressure was conceptualised as the degree of spatiotemporal constraint imposed on the attacking team during defensive sequences and was quantified using a composite index derived from six spatial tactical variables. Ordered logistic regression was used to examine the association between DPI and defensive outcomes, while one-way ANOVA was used to assess differences in DPI across running score. Higher DPI values were significantly associated with more favourable defensive outcomes (p < 0.001), providing preliminary evidence that the index captures meaningful sequence-level variation in defensive pressure. However, this finding should be interpreted with caution because the proportional odds assumption showed evidence of deviation. By contrast, DPI did not vary significantly across running score, suggesting that defensive pressure at the sequence level was more closely related to local spatial tactical organization than to scoreline alone. Overall, the DPI provides a transparent and replicable framework for quantifying defensive pressure in professional football. Future research should further evaluate the index using broader datasets and modelling approaches that account for the hierarchical structure of match data.

1 Introduction

Football action is structured around the collective objectives of two teams, each aiming to score more goals than the opposing squad. In this sense, the outcome of a match represents the ultimate social performance indicator in the sport (1). Within this competitive context, offensive and defensive processes are inherently interdependent, as each team continually seeks to overcome the opponent's efforts while preventing goalscoring opportunities. As an interactive team sport, football performance emerges not only from individual technical actions but also from the dynamic, intelligent coordination of players within spatial and temporal constraints (2–4). From a temporal perspective, football matches can be conceptualized as a continuous interaction of four interrelated phases: organised attack, organised defense, offensive transition following ball recovery, and defensive transition following ball loss (5).

These phases capture the evolving tactical structure of the game and provide a useful framework for analyzing team behaviors. With the increasing availability of tracking data, performance analysis in football has shifted towards more objective and finer grained approaches (6, 7). Tracking data enable the quantification of collective behaviors and spatial interactions between teams, providing new opportunities to analyze tactical organization and decision making processes (8, 9). Although substantial research has focused on attacking behaviors (10), physical characteristics (11–13), and technical actions (14, 15), defensive processes particularly during organised defensive phases and transitions remain comparatively underexplored. Little is known about how defensive pressure emerges and is maintained over time, as much of the existing literature assesses defensive performance using indirect or outcome based indicators.

Traditional defensive metrics, such as tackles, interceptions, fouls, or goals conceded, fail to capture the dynamic and collective nature of defensive behaviors (16, 17). These indicators do not adequately represent defensive pressure, which reflects both the immediate constraints imposed on the ball carrier and the broader spatial organization of the defending team (18, 19). Moreover, previous studies have examined defensive behaviors from multiple perspectives. For instance, some research has shown that teams adapt their defensive strategies by modifying spatial occupation and player synchronization (20), while different tactical formations (e.g., 4-4-2 vs. 5-3-2) can influence defensive organization and pressing behaviors (21). Other studies have focused on specific defensive strategies such as high pressing, counter pressing, and ball recovery dynamics (22–24), and Castellano and Pic (25) highlighted the distinction between high pressure and deep defensive strategies, while Santos et al. (26) demonstrated how pressing zones vary across tactical contexts.

As a result, there may be growing interest in developing more refined approaches to quantify defensive pressure using positional data. Several attempts have also been made to operationalize defensive pressure. Tenga et al. (27) categorized pressure into different qualitative levels, while Fernandez-Navarro et al. (28) quantified pressure using distances between attackers and defenders. More recently, studies using tracking data have emphasized the importance of spatial compactness and collective positioning in shaping defensive effectiveness (29). Despite these contributions, the concept of defensive pressure remains inconsistently defined, and there is still a lack of unified, replicable frameworks that integrate multiple spatial tactical variables into a comprehensive measure (18). In this sense, recent research has further emphasized the importance of spatial context in shaping defensive outcomes. For example, Jamil et al. (30) demonstrated that the effectiveness of organised pressing is strongly influenced by the location of ball recovery, suggesting that the place where possession is regained plays a critical role in determining subsequent defensive success.

Building on this perspective, the present study extends previous work by integrating multiple spatial tactical dimensions into a composite index (DPI) derived from tracking data, thereby providing a more comprehensive representation of defensive pressure beyond single variable approaches.

Moreover, existing research presents a degree of conceptual ambiguity regarding defensive pressure. In some studies, pressure is treated as a tactical strategy (e.g., pressing systems), while in others it is described as an emergent property reflecting intensity or spatial constraint (19). This lack of conceptual clarity limits the comparability of findings and highlights the need for more systematic approaches to defining and measuring defensive pressure.

Importantly, most empirical studies on defensive behaviors have been conducted in European leagues, with relatively limited research focusing on the Chinese Super League (CSL) (31, 32). Although some studies have examined defensive playing styles in the CSL, phase-level analyses based on tracking data, particularly those addressing defensive pressure, remain scarce. Consequently, the spatial characteristics and determinants of defensive pressure in this context remain poorly understood. In addition, while prior research has identified various defensive patterns and strategies, it remains unclear which specific spatial mechanisms most strongly contribute to the emergence of defensive pressure during defensive sequences. Specifically, the relative importance of factors such as regain location, defensive line positioning, and local player density has not been systematically examined within a unified analytical framework.

Therefore, to address these gaps, the present study focuses on the Chinese Super League and adopts a phase level approach based on positional tracking data. By constructing a composite Defensive Pressure Index (DPI) and evaluating its associations with defensive outcomes and match context, this study aims to provide a more transparent and replicable framework for analyzing defensive pressure. Specifically, the aims of this study were 1) to construct a tracking-data-based Defensive Pressure Index (DPI), 2) to evaluate whether the DPI is meaningfully associated with defensive outcomes, and 3) to examine whether the DPI varies according to the running score.

2 Methods

2.1 Sample and variables

This study adopted an observational, retrospective, and analytical design based on positional tracking data from official Chinese Super League matches. A total of 2,032 defensive sequences from 30 matches in the 2018 Chinese Super League (CSL) season, involving 16 participating teams, could be analysed. All data were collected through synchronized position tracking, event annotation and systematic recording. Each observation unit was defined as a defensive sequence, beginning when the opposing team gained possession of the ball and ending when the defending team regained possession, or the attacking team lost control of the ball.

To investigate the characteristics of defensive pressure, only organised attacking sequences were included in the analysis, that is, phases during which the attacking team maintained purposeful possession through a series of technical actions. Following the definitions proposed by (28, 33), an attacking team was considered to be in possession under either of the following conditions: the attacking team completed at least three consecutive passes during the same attacking phase, and the attacking player performed a dribbling action lasting at least three seconds. These criteria were adopted to exclude unintentional or transient possessions, such as single passes, unforced errors, or incidental ball contacts.

Figure 1

Figure 2

2.2 Procedure

The data for this study were generated using the Amisco® tracking system (Amisco®, Nice, France). Written permission from Amisco was obtained before the data collection. This system can precisely track the ball and all players throughout the entire match and generate a two dimensional animation model of player movements, enabling the quantitative analysis of both team and individual performances. All raw data from each game were stored in the computer system for subsequent post-match analysis. The data extraction process was conducted using the Amisco Viewer software (Figure 3). The validity and reliability of this system have been confirmed in previous studies (34, 35).

Figure 3

To ensure consistency between tracking data and defensive sequence event records, this study implemented a systematic data validation procedure. Two trained observers participated in the event annotation process. Prior to formal coding, both observers underwent systematic training to familiarize themselves with the tracking interface, coding procedures, and the operational definitions of all variables.

Following training, reliability testing was conducted to assess both intra-observer and inter-observer consistency. A randomly selected subsample representing 10% of the total dataset (n = 203 defensive sequences) was used for repeated coding. For intra-observer reliability, one observer re-coded the selected sequences after a two week interval. For inter-observer reliability, both observers independently coded the same subsample, and their coding results were subsequently compared.

For categorical variables, percentage agreement and Cohen's kappa coefficient were calculated. The inter observer agreement for defensive sequence outcomes was 91% (36), indicating high consistency between different observers. The corresponding Cohen's kappa coefficient was 0.78, indicating “good” agreement according to (37). The intra observer agreement (consistency of repeated assessments by the same observer) was 88%, with a Cohen's kappa of 0.74, demonstrating good reliability within observers.

For continuous variables (DPGZ, DPPBND, DNUM, PL, PN, and DLAODGL), reliability was assessed using intraclass correlation coefficients (ICC), Pearson correlation coefficients (r), and the typical error of measurement (TEM), following the methodology proposed by (38, 39). ICC values were calculated using a two-way mixed effects model for absolute agreement. These metrics collectively demonstrated strong repeatability and measurement precision. All reliability results are presented in Table 2. Overall, the validation procedures ensured the dataset was sufficiently reliable, providing a robust foundation for subsequent quantitative analysis.

2.3 Data processing

All variables were standardised using z-score normalisation (mean = 0, SD = 1) to remove differences in scale and ensure comparability across measures (40).

Prior to index construction, each variable was aligned according to its theoretical relationship with defensive pressure. Variables representing greater defensive intensity or spatial constraint (e.g., defensive density and territorial compression) contributed positively to the index, whereas variables associated with reduced defensive pressure (e.g., greater distance to the ball carrier or longer attacking sequences) contributed negatively. Accordingly, higher DPI values consistently indicate greater defensive pressure.

2.4 Variable definition and DPI construction

The Defensive Pressure Index (DPI) was defined as a composite indicator to quantify the degree of spatiotemporal constraint imposed by the defending team during a defensive sequence. As summarized in Table 1, the index was constructed from six spatial tactical variables: defensive possession gain zone (DPGZ), distance between the ball carrier and the nearest defender (DPPBND), number of defenders located around the ball carrier (DNUM), pass length of the final attacking pass before sequence termination (PL), number of attacking passes completed before sequence termination (PN), and distance from the least advanced outfield defender to the own goal line (DLAODGL).

These variables capture complementary aspects of defensive pressure, including territorial regain location, immediate pressure on the ball carrier, local defensive density, opponent progression patterns, attacking circulation prior to interruption, and defensive line depth. Accordingly, the DPI was designed to represent defensive pressure as a multidimensional, sequence level construct.

An equal weight formulation was adopted to ensure transparency and interpretability. This approach avoids reliance on sample specific weighting schemes and is consistent with the exploratory PCA results, which did not support a sufficiently strong latent structure for data driven weighting.

The DPI for each defensive sequence was computed as follows, where higher DPI values indicate greater defensive pressure:

2.5 Statistical analysis

2.5.1 Overview

The statistical analysis was designed to evaluate the Defensive Pressure Index (DPI) as an integrated measure of defensive pressure. Accordingly, the analytical framework tested whether the DPI was associated with defensive outcomes and match context, with subsequent models specified at the level of the composite index rather than its individual component variables.

2.5.2 Exploratory principal component analysis

An exploratory principal component analysis (PCA) was performed to examine the structure of the six spatial tactical variables included in the DPI (41).

The Kaiser-Meyer-Olkin (KMO) statistic was used to assess sample suitability, and Bartlett's sphericity test was conducted to check the appropriateness of the correlation matrix for factor analysis. The KMO statistic suggested some limitations in sample adequacy, while the Bartlett test indicated significant correlations between the variables (42, 43). Factors were extracted based on the Kaiser criterion (eigenvalues > 1), and multiple factors were retained.

2.5.3 Ordinal logistic regression analysis of the defensive pressure index (DPI)

An ordinal logistic regression model (44) was specified to examine whether the Defensive Pressure Index (DPI) was associated with defensive outcomes. This approach was considered appropriate because defensive outcome was treated as an ordered categorical variable with three levels: Fail (F), Gain Possession (GP), and Continue Defensive Sequence (CDS), reflecting increasing levels of defensive effectiveness within a sequence.

The DPI was entered as the sole predictor to evaluate the association between the composite index and defensive outcomes at the sequence level. Parameter estimates were interpreted as odds ratios, indicating the change in the odds of achieving a higher level defensive outcome category for a one-unit increase in DPI. Model fit was summarized using McFadden's pseudo-R², and the proportional odds assumption was assessed using the Brant test prior to interpretation.

2.5.4 Match context analysis

To examine whether defensive pressure varied according to match context, a one-way analysis of variance (ANOVA) was conducted with running score (losing, drawing, winning) as the independent variable and the Defensive Pressure Index (DPI) as the dependent variable. post hoc pairwise comparisons were performed using Tukey's HSD test. Effect size was estimated using eta squared η² (45).

2.5.5 Software and significance level

All statistical analyses were performed in R version 4.5.2. Excel data were imported using readxl, and data processing and cleaning were conducted using dplyr and janitor. Descriptive statistics and reliability summaries were calculated using psych and DescTools. Ordinal logistic regression was performed using MASS, McFadden's pseudo-R² was calculated using pscl, and the proportional odds assumption was assessed using brant. Analysis of variance and Tukey's post hoc comparisons were conducted using base R functions. Figures were created using ggplot2 and fact extra. Statistical significance was set at p < 0.05.

3 Results

3.1 Exploratory structure of defensive variables (PCA)

As a supplementary exploratory check, a principal component analysis (PCA) was conducted on the six spatial tactical variables. Sampling adequacy was limited (KMO = 0.51), although Bartlett's test of sphericity was significant, χ²(15) = 344.37, p < 0.001.

Two components had eigenvalues greater than 1 and together explained 43.08% of the total variance (see Figure 4). However, the component structure did not reveal a clear low-dimensional pattern. Therefore, the PCA was interpreted for exploratory purposes only and was not used for variable reduction, weighting, or subsequent modelling. Detailed PCA outputs are provided in the Supplementary Materials.

Figure 4

3.2 Validation of the defensive pressure Index (DPI)

The ordinal logistic regression analysis (see Table 3) provided preliminary evidence of an association between the Defensive Pressure Index (DPI) and defensive outcomes. Higher DPI values were associated with a greater likelihood of achieving higher-level defensive outcomes. Specifically, a one-unit increase in DPI was associated with an increase in the odds of attaining a higher defensive outcome category (β = 0.298, SE = 0.019, z = 15.77, p < 0.001), corresponding to a 34.7% increase in odds (OR = 1.347, 95% CI [1.298, 1.398).

The model demonstrated modest explanatory power (McFadden's pseudo-R² = 0.0738), indicating that while DPI captures part of the variation in defensive outcomes, additional contextual and tactical factors are likely to play a role.

The proportional odds assumption was assessed using the Brant test and showed some evidence of deviation (p < 0.001), suggesting that the effect of DPI may not be entirely constant across outcome thresholds. Therefore, the results should be interpreted with appropriate caution.

Overall, these findings suggest that the DPI provides a useful sequence level summary of defensive pressure, while further validation using alternative modelling approaches is warranted.

3.3 Effect of running score on defensive pressure

To examine whether defensive pressure varied according to match context, a one-way analysis of variance (ANOVA) was conducted with running score (losing, drawing, winning) as the independent variable and the Defensive Pressure Index (DPI) as the dependent variable. Tukey's HSD test was used for post hoc comparisons, and effect size was estimated using eta squared (η²).

The analysis indicated that running score did not have a statistically significant effect on DPI, F(2, 2029) = 0.69, p = 0.504, η² = 0.001. Mean DPI values showed only minor differences across running score (see Table 4),and Tukey's HSD post hoc tests revealed no significant pairwise differences between conditions (p > 0.49).

These findings suggest that defensive pressure, as captured by the DPI, does not vary systematically across different running score at the sequence level (see Figure 5).

Figure 5

4 Discussion

The present study aimed to quantify defensive pressure using tracking data and to develop a composite measure, the Defensive Pressure Index (DPI), to capture its multidimensional nature in Chinese Super League (CSL) matches. Overall, the findings suggest that defensive pressure, as operationalised by the DPI, is not solely the result of isolated defensive actions but rather emerges from the collective spatial organization of the defending team. In this sense, the study shifts the analytical focus from individual spatial tactical variables to an integrated index, allowing defensive pressure to be represented as a more coherent and operational construct, which is important given the complexity of team sports.

Previous research has highlighted the importance of spatial tactical factors in shaping defensive pressure (18, 19, 46, 47). The present study extends this line of work by combining these dimensions into a single composite index, rather than analyzing them separately, offering a potentially more holistic and interpretable representation of defensive pressure in dynamic match situations.

A key finding of the study is that higher DPI values were associated with more favourable defensive outcomes. Sequences characterized by higher defensive pressure were more likely to correspond to higher level defensive outcome categories. This aligns with previous research indicating that defensive pressure can disrupt attacking organization and increase the likelihood of successful defensive sequences (18, 46). Importantly, these findings suggest that this relationship can be captured through an integrated index, providing preliminary support for the DPI as a sequence level summary of defensive pressure. However, the proportional odds assumption showed evidence of deviation, indicating that the association between DPI and defensive outcomes may not be constant across outcome thresholds. This issue is addressed further in the limitations section.

In contrast, the running score had no statistically significant effect on DPI, and the effect size was negligible. This suggests that, within this sample, the defensive pressure reflected by DPI remained largely stable across different running score (losing, drawing, and winning). This could indicate that the structural components of DPI are more closely tied to the collective organization of the defense, rather than to the immediate scoreline context. While no statistically significant differences were found, a faint trend may still be observed. One possible explanation is that DPI primarily reflects local spatiotemporal interactions such as distance from the ball carrier, local density, recovery positions, and defensive line organization factors more directly influenced by positional constraints than by the broader match context. In this sense, the scoreline may influence tactical intentions, but these influences may not necessarily translate into consistent changes in defensive pressure at the sequence level. This interpretation is in line with recent studies that have highlighted the importance of spatial context in defensive performance. For example, Jamil et al. (30) showed that the effectiveness of organised pressing is influenced by spatial factors such as ball recovery location. The present findings seem consistent with this perspective, suggesting that defensive pressure is fundamentally shaped by spatial tactical elements.

At the same time, the nonsignificant effect of running score should not be interpreted as evidence that contextual factors are irrelevant. Rather, it suggests that the running score alone may be too broad a variable to capture the complexity underlying defensive behaviors. Defensive pressure is likely shaped by a wider array of situational and tactical factors that were not explicitly modelled in the present study.

This study also provides some initial insight into defensive behaviors in the CSL. Compared with patterns reported in European leagues, defensive pressure in the CSL sample appears to occur more frequently in deeper areas of the pitch, potentially suggesting a more conservative defensive approach. However, as no direct cross league comparison was conducted, this interpretation remains tentative and should be examined more systematically in future studies.

5 Limitations

The present study has several limitations that should be considered when interpreting the findings. First, the analyses were conducted at the level of individual defensive sequences, which were nested within matches and teams, but the models did not explicitly account for this hierarchical data structure. As a result, potential non independence among observations may have influenced standard errors and significance estimates. Future research should address this issue by using mixed effects or other hierarchical modelling approaches to better capture variation across teams and matches.

The statistical evaluation of the DPI should also be interpreted with caution. Although the index was significantly associated with defensive outcomes, the explanatory power of the model was modest, indicating that defensive performance is likely influenced by additional contextual and tactical factors not included in the present analysis. Moreover, the proportional odds assumption showed evidence of deviation, suggesting that the relationship between defensive pressure and outcomes may differ across outcome thresholds. This highlights the need for more flexible modelling approaches in future studies.

Several data related limitations should also be acknowledged. The sample was restricted to 30 matches from a single league, which may limit the generalisability of the findings to other competitions or playing contexts. In addition, the analysis focused only on organised attacking sequences with a minimum duration and multiple passes, meaning that transitional or short duration defensive situations were not captured. Finally, given the observational design of the study, the findings should be interpreted as associative rather than causal.

6 Future research

Future research should extend this framework by incorporating a broader range of contextual variables. Factors such as coaching strategy, tactical instructions, match period, fatigue, opponent quality, and relative team strength may all influence the application of defensive pressure.

In addition, underlying tactical structures may play an important role. Previous studies have shown that formations and positional organization can influence defensive behaviors and pressing dynamics (21). As these factors were not explicitly modelled in the present study, future research should incorporate formation and positional data to better understand their contribution to defensive pressure.

Methodologically, the use of mixed effects models, longitudinal designs, and simulation based approaches using tracking data may provide further insight into how defensive pressure evolves across different match contexts.

7 Conclusion

Overall, the findings suggest that defensive pressure, as captured by the Defensive Pressure Index, can be understood as a multidimensional construct emerging from collective spatial organization rather than isolated individual actions. The DPI was associated with defensive outcomes, providing preliminary evidence that it may serve as a useful sequence level summary of defensive pressure.

By contrast, DPI showed limited variation across running score, suggesting that running score alone may not adequately capture the contextual complexity underlying defensive behavior. Despite its limitations, the present study provides an operational and replicable framework for quantifying defensive pressure and offers a foundation for future research into the contextual and tactical determinants of defensive organization in professional football.

References

• 1.

  KempeMMemmertD. “Good, better, creative”: the influence of creativity on goal scoring in elite soccer. J Sports Sci. (2018) 36(21):2419–23. 10.1080/02640414.2018.1459153

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 2.

  ParlebasP. The universals of games and sports. Front Psychol Vol. (2020):11–2020. 10.3389/fpsyg.2020.593877

  • CrossRef
  • Google Scholar

• 3.

  Martínez-SantosRNubiolaJ. Signe, science et jeux sportif: esquisse de sémiotricité triadique. Semiotica. (2022) 2022(248):77–103. 10.1515/sem-2022-0069

  • CrossRef
  • Google Scholar

• 4.

  ParlebasP. Semiotricité et corps en jeu. Semiotica. (2022) 2022(248):9–35. 10.1515/sem-2022-0081

  • CrossRef
  • Google Scholar

• 5.

  DuarteRAraújoDCorreiaVDavidsKMarquesPRichardsonMJ. Competing together: assessing the dynamics of team-team and player-team synchrony in professional association football. Hum Mov Sci. (2013) 32(4):555–66. 10.1016/j.humov.2013.01.011

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 6.

  MemmertDLemminkKAPMSampaioJ. Current approaches to tactical performance analyses in soccer using position data. Sports Med. (2017) 47(1):1–10. 10.1007/s40279-016-0562-5

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 7.

  GoesFRMeerhoffLABuenoMJORodriguesDMMouraFABrinkMSet alUnlocking the potential of big data to support tactical performance analysis in professional soccer: a systematic review. Eur J Sport Sci. (2021b) 21(4):481–96. 10.1080/17461391.2020.1747552

  • CrossRef
  • Google Scholar

• 8.

  RavéGGranacherUBoullosaDHackneyACZouhalH. How to use global positioning systems (GPS) data to monitor training load in the “real world” of elite soccer. Front Physiol Vol. (2020):11–2020. 10.3389/fphys.2020.00944

  • CrossRef
  • Google Scholar

• 9.

  KovalchikSA. Player tracking data in sports. Annu Rev Stat Appl. (2023) 10:677–97. 10.1146/annurev-statistics-033021-110117

  • CrossRef
  • Google Scholar

• 10.

  GoesFRBrinkMSElferink-GemserMTKempeMLemminkKAPM. The tactics of successful attacks in professional association football: large-scale spatiotemporal analysis of dynamic subgroups using position tracking data. J Sports Sci. (2021a) 39(5):523–32. 10.1080/02640414.2020.1834689

  • CrossRef
  • Google Scholar

• 11.

  Martinez-SantosRCastilloDLos ArcosA. Sprint and jump performances do not determine the promotion to professional elite soccer in Spain, 1994–2012. J Sports Sci. (2016) 34(24):2279–85. 10.1080/02640414.2016.1190460

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 12.

  Los ArcosACastilloDMartínez-SantosR. Influence of initial performance level and tactical position on the aerobic fitness in soccer players after preseason period. Sci Med Football. (2018) 2(4):294–8. 10.1080/24733938.2018.1451652

  • CrossRef
  • Google Scholar

• 13.

  CastellanoJHuarteXCasamichanaD. Match physical performance profiles in professional football: a comparative analysis among players’ positions in the European five top leagues. Int J Perform Anal Sport. (2025) 25(1):108–28. 10.1080/24748668.2024.2393034

  • CrossRef
  • Google Scholar

• 14.

  BlackGMGabbettTJNaughtonGAMcLeanBD. The effect of intense exercise periods on physical and technical performance during elite Australian football match-play: a comparison of experienced and less experienced players. J Sci Med Sport. (2016) 19(7):596–602. 10.1016/j.jsams.2015.07.007

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 15.

  VellaAClarkeACKemptonTRyanSHoldenJCouttsAJ. Technical involvements and pressure applied influence movement demands in elite Australian football match-play. Sci Med Football. (2022) 6(2):228–33. 10.1080/24733938.2021.1942537

  • CrossRef
  • Google Scholar

• 16.

  GaviãoLOSant’AnnaAPAlves LimaGBde Almada GarciaPA. Evaluation of soccer players under the moneyball concept. J Sports Sci. (2020) 38(11-12):1221–47. 10.1080/02640414.2019.1702280

  • CrossRef
  • Google Scholar

• 17.

  LepschyHWäscheHWollA. Success factors in football: an analysis of the German bundesliga. Int J Perform Anal Sport. (2020) 20(2):150–64. 10.1080/24748668.2020.1726157

  • CrossRef
  • Google Scholar

• 18.

  AndrienkoGAndrienkoNBudziakGDykesJFuchsGvon LandesbergerTet alVisual analysis of pressure in football. Data Min Knowl Discov. (2017) 31(6):1793–839. 10.1007/s10618-017-0513-2

  • CrossRef
  • Google Scholar

• 19.

  ForcherLForcherLAltmannSJekaucDKempeM. The keys of pressing to gain the ball—characteristics of defensive pressure in elite soccer using tracking data. Sci Med Football. (2024c) 8(2):161–9. 10.1080/24733938.2022.2158213

  • CrossRef
  • Google Scholar

• 20.

  FreitasRVolossovitchAAlmeidaCHVleckV. Elite-level defensive performance in football: a systematic review. Ger J Exerc Sport Res. (2023) 53(4):458–70. 10.1007/s12662-023-00900-y

  • CrossRef
  • Google Scholar

• 21.

  LowBReinRSchwabSMemmertD. Defending in 4–4-2 or 5-3-2 formation? Small differences in footballers’ collective tactical behaviours. J Sports Sci. (2022) 40(3):351–63. 10.1080/02640414.2021.1993655

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 22.

  AlmeidaCHFerreiraAPVolossovitchA. Effects of match location, match status and quality of opposition on regaining possession in UEFA champions league. J Hum Kinet. (2014) 41:203–14. 10.2478/hukin-2014-0048

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 23.

  VogelbeinMNoppSHökelmannA. Defensive transition in soccer—are prompt possession regains a measure of success? A quantitative analysis of German fußball-bundesliga 2010/2011. J Sports Sci. (2014) 32(11):1076–83. 10.1080/02640414.2013.879671

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 24.

  HarperDJSandfordGNClubbJYoungMTabernerMRhodesDet alElite football of 2030 will not be the same as that of 2020: what has evolved and what needs to evolve?Scand J Med Sci Sports. (2021) 31(2):493–4. 10.1111/sms.13876

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 25.

  CastellanoJPicM. Identification and preference of game styles in LaLiga associated with match outcomes. Int J Environ Res Public Health. (2019) 16(24):5090. 10.3390/ijerph16245090

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 26.

  SantosPLago-PeñasCGarcía-GarcíaO. The influence of situational variables on defensive positioning in professional soccer. Int J Perform Anal Sport. (2017) 17(3):212–9. 10.1080/24748668.2017.1331571

  • CrossRef
  • Google Scholar

• 27.

  TengaAHolmeIRonglanLTBahrR. Effect of playing tactics on goal scoring in Norwegian professional soccer. J Sports Sci. (2010) 28(3):237–44. 10.1080/02640410903502774

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 28.

  Fernandez-NavarroJRuiz-RuizCZubillagaAFraduaL. Tactical variables related to gaining the ball in advanced zones of the soccer pitch: analysis of differences among elite teams and the effect of contextual variables. Front Psychol Vol. (2020):10–2019. 10.3389/fpsyg.2019.03040

  • CrossRef
  • Google Scholar

• 29.

  ForcherLForcherLAltmannSJekaucDKempeM. Is a compact organization important for defensive success in elite soccer?—analysis based on player tracking data. Int J Sports Sci Coach. (2024b) 19(2):757–68. 10.1177/17479541231172695

  • CrossRef
  • Google Scholar

• 30.

  JamilMMcElduffCAdsheadLWelshJBeatoMEz-ziziA. A spatial analysis of ball recovery locations, ball recovery methods and eventual possession outcomes in elite women’s football using a Bayesian network approach. Int J Perform Anal Sport. (2025) 1–16. 10.1080/24748668.2025.2557776

  • CrossRef
  • Google Scholar

• 31.

  RuanLGeHGómezM-ÁShenYGongBCuiY. Analysis of defensive playing styles in the professional Chinese football super league. Sci Med Football. (2023) 7(3):279–87. 10.1080/24733938.2022.2099964

  • CrossRef
  • Google Scholar

• 32.

  WangZFangZDaiRNiuZLiuH. Match performance analysis on the Chinese football association super league: a systematic review. Int J Sports Sci Coach. (2025) 20(2):812–26. 10.1177/17479541251314185

  • CrossRef
  • Google Scholar

• 33.

  ForcherLBeckmannTWohakORomeikeCGrafFAltmannS. Prediction of defensive success in elite soccer using machine learning—tactical analysis of defensive play using tracking data and explainable AI. Sci Med Football. (2024a) 8(4):317–32. 10.1080/24733938.2023.2239766

  • CrossRef
  • Google Scholar

• 34.

  CarlingCBloomfieldJNelsenLReillyT. The role of motion analysis in elite soccer: contemporary performance measurement techniques and work rate data. Sports Med. (2008) 38(10):839–62. 10.2165/00007256-200838100-00004

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 35.

  CastellanoJAlvarez-PastorDBradleyPS. Evaluation of research using computerised tracking systems (amisco and prozone) to analyse physical performance in elite soccer: a systematic review. Sports Med. (2014) 44(5):701–12. 10.1007/s40279-014-0144-3

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 36.

  GastDLLedfordJR. Single Subject Research Methodology in Behavioral Sciences. New York, NY: Routledge (2010).

  • Google Scholar

• 37.

  AshbyD. Practical statistics for medical research. Douglas G. Altman, chapman and Hall, London, 1991. No. Of pages: 611. Price: £32.00. Stat Med. (1991) 10(10):1635–6. 10.1002/sim.4780101015

  • CrossRef
  • Google Scholar

• 38.

  HopkinsWG. Measures of reliability in sports medicine and science. Sports Med. (2000) 30(1):1–15. 10.2165/00007256-200030010-00001

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 39.

  HopkinsWG. Spreadsheets for analysis of validity and reliability. Sportsci. (2017) 19:36–38.

  • Google Scholar

• 40.

  JamesG. An Introduction to Statistical Learning: With Applications in R. Cham, Springer (2013).

  • Google Scholar

• 41.

  JolliffeI. Principal component analysis. In: Encyclopedia of statistics in behavioral science (Chichester, UK: John Wiley & Sons) (2005). 10.1002/0470013192.bsa501

  • CrossRef
  • Google Scholar

• 42.

  BartlettMS. A note on the multiplying factors for Various χ2 approximations. J R Stat Soc B (Methodological). (1954) 16(2):296–8. 10.1111/j.2517-6161.1954.tb00174.x

  • CrossRef
  • Google Scholar

• 43.

  KaiserHF. The application of electronic computers to factor analysis. Educ Psychol Meas. (1960) 20(1):141–51. 10.1177/001316446002000116

  • CrossRef
  • Google Scholar

• 44.

  McCullaghP. Regression models for ordinal data. J R Stat Soc B (Methodological). (1980) 42(2):109–27. 10.1111/j.2517-6161.1980.tb01109.x

  • CrossRef
  • Google Scholar

• 45.

  CohenJ. Statistical Power Analysis for the Behavioral Sciences. 2nd ed., Mahwah, NJ: Lawrence Erlbaum Associates (1988).

  • Google Scholar

• 46.

  WrightCAtkinsSPolmanRJonesBSargesonL. Factors associated with goals and goal scoring opportunities in professional soccer. Int J Perform Anal Sport. (2011) 11(3):438–49. 10.1080/24748668.2011.11868563

  • CrossRef
  • Google Scholar

• 47.

  LowBReinRRaabeDSchwabSMemmertD. The porous high-press? An experimental approach investigating tactical behaviours from two pressing strategies in football. J Sports Sci. (2021) 39(19):2199–210. 10.1080/02640414.2021.1925424

  • Pubmed Abstract
  • CrossRef
  • Google Scholar