Comparative analysis of key performance indicators in Euroleague and national basketball leagues

Abstract

Background: Understanding the performance metrics that underpin team success in the dynamic professional basketball world is crucial. This study originated in the context of increasing academic and professional interest in performance analytics, focusing on how teams in elite leagues, such as the Euroleague and their respective national leagues, adapt and perform based on specific key performance indicators (KPIs). Purpose: The primary objective of this research is to bridge the existing bibliographic gap by comparing the effectiveness of various KPIs in predicting match outcomes in both the Euroleague and National Basketball Leagues. This comparison aims to identify how strategic adaptations and performance measures differ according to the unique demands and styles of the respective competitions. Methodology: The study utilized two main datasets: one encompassing all Euroleague 2022–23 matches and the other compiling cumulative statistics from Euroleague teams over three seasons. Machine learning techniques, including Random Forest, Logistic Regression, and Support Vector Machines, were employed along with the Boruta algorithm for feature selection to enhance predictive accuracy and SHapley Additive exPlanations (SHAP) for the interpretability of the model output. Results: The analysis identified that certain KPIs, such as the effective field goal percentage, defensive ratings, and assists-to-turnover ratio, vary significantly in their impact on game outcomes between Euroleague and National League games. These variations imply that teams may need to tailor their strategies depending on the league in which they play. Conclusions: This study significantly advances the field of sports analytics by providing a detailed comparative analysis of basketball performance metrics across two competitive settings. It offers practical insights that can help coaches and analysts optimize team performance and strategic planning. Moreover, sophisticated data analysis techniques have allowed for a deeper understanding of the complex dynamics that influence basketball game outcomes, thereby making a significant contribution to the literature and practice of sports performance analysis.

Headnote

Abstract

Background: Understanding the performance metrics that underpin team success in the dynamic professional basketball world is crucial. This study originated in the context of increasing academic and professional interest in performance analytics, focusing on how teams in elite leagues, such as the Euroleague and their respective national leagues, adapt and perform based on specific key performance indicators (KPIs). Purpose: The primary objective of this research is to bridge the existing bibliographic gap by comparing the effectiveness of various KPIs in predicting match outcomes in both the Euroleague and National Basketball Leagues. This comparison aims to identify how strategic adaptations and performance measures differ according to the unique demands and styles of the respective competitions. Methodology: The study utilized two main datasets: one encompassing all Euroleague 2022–23 matches and the other compiling cumulative statistics from Euroleague teams over three seasons. Machine learning techniques, including Random Forest, Logistic Regression, and Support Vector Machines, were employed along with the Boruta algorithm for feature selection to enhance predictive accuracy and SHapley Additive exPlanations (SHAP) for the interpretability of the model output. Results: The analysis identified that certain KPIs, such as the effective field goal percentage, defensive ratings, and assists-to-turnover ratio, vary significantly in their impact on game outcomes between Euroleague and National League games. These variations imply that teams may need to tailor their strategies depending on the league in which they play. Conclusions: This study significantly advances the field of sports analytics by providing a detailed comparative analysis of basketball performance metrics across two competitive settings. It offers practical insights that can help coaches and analysts optimize team performance and strategic planning. Moreover, sophisticated data analysis techniques have allowed for a deeper understanding of the complex dynamics that influence basketball game outcomes, thereby making a significant contribution to the literature and practice of sports performance analysis.

Keywords: sports analytics, performance analysis, basketball metrics, machine learning, SHAP

Introduction

In recent years, there has been a significant surge in interest regarding the examination of athletic performance. This interest spans both academic research and its real-world implementation in sports (Plakias et al., 2023a; Plakias et al., 2023b). In the realm of professional basketball, success is often measured by a multifaceted evaluation of a team's performance. Coaches, analysts, and enthusiasts scrutinize various metrics and statistics to understand the dynamics of the game and gauge a team's proficiency (Izzo et al., 2023). The Euroleague showcases the pinnacle of European basketball, featuring teams that have honed their skills to compete on the grandest stage (Bourdas et al., 2022). The high intensity and caliber of the games in this elite competition render it an invaluable source of insight into the performance of top-tier basketball teams. However, to comprehend a team's overall capabilities and strategies fully, we must also examine its performance within the context of its domestic league.

A focal point within performance analysis is the study of performance indicators (PIs), which play a pivotal role in offering valuable insights (Hughes & Bartlett, 2002; Kapsalis et al., 2023). However, to derive more useful and reliable conclusions, researchers and team coaching staff focus on the use of KPIs, which are the metrics that contribute most significantly to success (Butterworth et al., 2013; Plakias et al., 2024). Various methods have been used to identify KPIs in basketball. Mikołajec et al. (2021) used econometrics and modeling prediction techniques, utilizing descriptive statistics, variable correlation comparisons, and multiple regression analyses in EuroLeague matches from 2003 to 2016. Lampis et al. (2023) employed three algorithms (logistic regression (LR), random forests (RF), and extreme gradient boosting trees) and predictive ensemble methods analyzing data from 5214 matches across four different European basketball tournaments (EuroLeague, EuroCup, Greek Basket League, and Spanish Liga ACB) for the period 2013-2018. Li et al. (2023) applied K-Means clustering and C5.0 decision trees for games in the Chinese Basketball Association (CBA). Zhou et al. (2024) used multiple linear regression (MLR) and quantile regression (QR) analysis for NBA games. While these methods provide significant insights, they cannot offer holistic interpretability and precise estimation of each variable's influence on match outcomes.

In the scholarly literature on basketball, several studies have addressed variances in performance metrics across different competitions, revealing distinct patterns and factors influencing team performance. For instance, Mandić et al. (2019) compared the NBA and Euroleague, emphasizing differences in game pace and possessions per game due to divergent athletic and tactical focuses. Similarly, Paulauskas et al. (2018) highlighted significant disparities in performance statistics related to body size and basketball skills between Euroleague and NBA players during the EuroBasket 2015. Further emphasizing the contextual influence on basketball performance, Muro et al. (2020) discussed how competition levels affect various aspects of basketball events, including intensity and risk factors. This perspective is complemented by studies like Ermiş et al. (2019), who pointed out performance variations between Euroleague players' contributions to their national teams, suggesting differing demands and expectations in international versus domestic settings. The impact of league-specific characteristics on PIs was also explored by Tormo et al. (2015), who examined free throw incidence across national and European leagues, and Oliveira-Da-Silva et al. (2013) who identified higher intensity demands in the Euroleague compared to domestic leagues for elite female basketball players. Despite these valuable insights, there appears to be a bibliographic gap specifically concerning the differences in KPIs for men's basketball teams when comparing their performances in the Euroleague with their respective games in national competitions.

Understanding the performance metrics that underpin success in professional basketball is crucial, particularly in a competitive landscape where the Euroleague and national leagues demand distinct strategic adaptations. The necessity for this research stems from the observed bibliographic gap in comparative studies focusing on KPIs across different competitive contexts. The context of this work is grounded in the growing interest in performance analytics within elite basketball competitions, where understanding the differential impact of KPIs can significantly influence coaching strategies and game outcomes. Given the current research landscape, which largely focuses on performance metrics within specific basketball leagues without extensive comparative analyses across different competitive environments, our study hypothesized that KPIs influencing victories in the Euroleague might significantly differ when these teams compete in their respective domestic leagues. This hypothesis is based on the premise that strategic adaptations and player performances are distinctly tailored to the unique demands and styles of each competition. Therefore, the primary aim of this article is to bridge the bibliographic gap by identifying which KPIs are most influential in winning games in the Euroleague and then comparing these KPIs to performances in national leagues to discern any significant differences. By doing so, this study provides a comprehensive insight into the tactical and strategic adaptations required for success in different competitive settings, thereby contributing to both the academic field of sports analytics and practical coaching methodologies.

This manuscript distinguishes itself with several unique contributions to the field of sports analytics. Firstly, it conducts a comprehensive KPI analysis across two distinct competitive contexts, systematically identifying and comparing the KPIs for Euroleague teams within both the Euroleague and their domestic competitions. This approach provides novel insights into how different playing environments influence team performance strategies, marking a significant advancement in understanding strategic adaptations in professional basketball. To explore the differential impact of KPIs across the Euroleague and national basketball leagues, we employed a multifaceted research approach incorporating both traditional statistical analyses and advanced machine learning techniques. Initially, we collected comprehensive datasets encompassing all matches from the Euroleague 2022-23 season and cumulative statistics from Euroleague teams over three seasons, including their performances in domestic leagues. Secondly, the study utilizes advanced machine learning (ML) techniques to enhance predictive accuracy and interpretability. By integrating RF, LR, and Support Vector Machines (SVMs), alongside the Boruta feature selection method and SHAP model, the research not only pinpoints the most crucial KPIs but also elucidates their impacts on match outcomes in a detailed and comprehensible manner. In particular, for interpretability, we utilized SHAP values, which allowed us to quantify the impact of each feature on the model's predictions. This step was crucial for understanding the underlying relationships within the data and provided actionable insights into the most influential KPIs. Finally, the empirical validation of theoretical predictions through rigorous machine learning processes and robust evaluation metrics such as accuracy, recall, precision, F1-score, and ROC curves, ensures that the study's findings are both theoretically sound and practically viable. These attributes underscore the manuscript's substantial contribution to existing literature, offering innovative perspectives and methods that could profoundly impact the application and understanding of performance analytics in professional basketball.

Material & methods

Sample

Two distinct datasets were utilized for this study. The first dataset encompassed all Euroleague 2022-23 matches, spanning the regular season and playoffs. It consisted of two observations per match, one for each team, resulting in a total of 656 observations extracted from 328 matches. Each observation included the match outcome (win or loss) along with the 80 variables derived from the Instat Basketball analysis platform, offering team statistics. The table in Appendix A shows the 80 variables together with their respective definitions. This initial dataset aimed to identify the most influential among the 80 variables concerning the match's result.

The second dataset focused on the cumulative statistics, averaged per match, of the teams engaged in the Euroleague over the last three seasons (2020-21, 2021-22, 2022-23), in both their Euroleague and domestic league fixtures. The table in Appendix B shows the 51 teams that participated in the Euroleague in the 3 aforementioned seasons (18 + 15 + 18). However, Olympiacos 2020-21 season was excluded from this group, as it did not participate in the 1st division of the Greek league.

Consequently, the second dataset comprised 50 observations, each corresponding to a different team. Each observation in the second dataset featured 160 variables, with 80 related to Euroleague matches and another 80 tied to National league matches. From this second dataset, only the variables determined to significantly influence a team's success were considered. The second dataset was employed to conduct a comparative analysis between KPIs in the Euroleague and their corresponding metrics in the national leagues.

Procedure

Data was acquired from Instat Basketball in distinct Excel sheets, one for each team (access on 15/10/2023 upon request). Instat Basketball facilitates data export in XLS format. Previous research has tested the reliability and validity of the data contained in Instat Basketball, showing very high values (BustamanteSánchez et al., 2022). The authors subsequently merged these individual Excel sheets into a single comprehensive sheet, making essential adjustments to prepare the data for subsequent analysis.

Machine learning Analyses

Problem definition: In this study, our main goal was to develop an explainable machine learning approach that could identify essential informative factors influencing match outcome predictions. We also explored the impact of these factors on the model's output, with a specific focus on post hoc explainability. To accomplish this, we treated the task of predicting match outcomes as a binary classification problem. To be specific, we divided the observations into two categories: (i) the "win" group, comprising 328 observations where teams emerged victorious and (ii) the "lose" group, consisting of 328 observations where teams lost.

Feature Engineering: StandardScaler library was applied for data normalization, which is a crucial step in ensuring fair comparisons between different features by scaling them to a standard range. Normalization enhances the performance and reliability of machine learning models. Subsequently, we utilized the Boruta algorithm for feature selection (FS), a powerful technique that helps identify the most relevant features for prediction. Boruta efficiently evaluates feature importance by comparing the importance of observed features against a shadow feature set. This rigorous selection process ensures that only the most informative features are considered, enhancing the accuracy and interpretability of our predictive model.

Learning process: In our learning process, we employed a diverse set of ML classifiers, including RF, LR, and SVM, each optimized with hyperparameters to enhance their predictive power. To rigorously evaluate the performance of these models, we adopted a stochastic validation strategy: 70% training/30% testing. To evaluate the models' effectiveness, we utilized a comprehensive set of performance metrics on the testing set. These metrics included accuracy, which measures the overall correctness of predictions, recall and precision, which assess the model's ability to capture relevant instances and minimize false positives respectively, and F1score, which balances the trade-off between precision and recall.

Additionally, we analyzed the confusion matrix, providing a detailed breakdown of true positive, true negative, false positive, and false negative predictions, offering insights into model performance across different classes. Furthermore, we utilized ROC curves, which graphically represent the model's ability to discriminate between positive and negative instances at various thresholds. This comprehensive evaluation approach ensured a thorough assessment of the models' predictive capabilities and robustness. Furthermore, we utilized accuracy, a widely adopted measure for assessing the overall performance of our models.

Interpretation: The SHAP model plays a crucial role in enhancing the interpretability of machine learning models. It quantifies the impact of each feature on a model's prediction, revealing intricate relationships within complex datasets based on game theory. By employing SHAP, we gain a deeper understanding of the importance of specific informative factors in predicting match outcomes (Moustakidis et al., 2023).

Statistical analyses

For the 18 resultant variables, Kolmogorov-Smirnov tests were conducted on the sample of 50 teams (second dataset) for the values of the PIs in both the Euroleague and their respective domestic leagues. Paired samples t-tests were applied to variables that exhibited a normal distribution in both the Euroleague and the domestic leagues. In cases where the data did not follow a normal distribution, the corresponding non-parametric Wilcoxon's signed-ranks test was employed.

All analyses were carried out using IBM SPSS statistical software (version 29.0). The significance level was set at p<0.05. Cohen's d was computed to indicate the effect size, which was categorized as follows: trivial (d=0.0 to 0.19), small (d=0.2 to 0.49), medium (d=0.5 to 0.79), large (d=0.8 to 1.29), and very large (d≥1.3).

Results

In this section are presented the selected factors from the Boruta FS algorithm, the testing performance metrics of the employed ML classifiers and the interpretation of the model output of the best performed ML classifier as well as the results of the statistics following the identification of the main contributing factors for the Euroleague and the domestic leagues.

Machine Learning

Table 1 demonstrates the 18 most informative factors after the implementation of the Boruta algorithm for FS. The factors were prioritized in order of importance, from highest to lowest. Table 1. Most informative identified factors.

Table 2 summarizes the testing performance results for the employed ML classifiers in the 18 most informative factors in our binary task. The best testing performance was achieved from the LR classifier. Specifically, LR achieved 94.92% accuracy, 95.83% recall, 93.88% precision and 94.85% f1-score. On the other hand, the lowest testing performance in this binary task was achieved by RF. In particular, RF achieved 91.88% accuracy, 90.62% recall, 92.55% precision and 91.58% f1-score. Figure 1 presents the ROC curve performance of the employed ML classifiers.

Figure 2 illustrates the influence of the most informative factors on the output of the best-performing model, which is the LR ML model. Figure 2a displays the most influential predictive factors in descending order, providing a top-down perspective. The colors represent the risk level associated with match observations, with red indicating high values and blue indicating low values. Specifically, high defensive rating values correspond to a higher likelihood of losing the match, establishing a clear association between defensive rating and match outcome. Additionally, high values in factors such as 'field goals made' exert a positive influence on the match's outcome, while low values in offensive rating, points, and effective field goal percentage are negatively associated with winning the match. Figure 2b depicts the average influence of the selected informative factors on the model's output magnitude. Notably, defensive rating, offensive rating, points, and effective field goal percentage significantly contribute to the model's output. Conversely, factors such as field goals made, rebounds, and contested field goals made have a moderate impact.

Statistics

Table 3 displays the means and standard deviations for the 18 variables, both for the Euroleague and the domestic leagues.

The Kolmogorov-Smirnov test revealed that out of the 36 variables used from the second dataset, only 4 (Points_per_possession_NAT, Points_per_possession_EURO, Defensive_rebounds_EURO, Offensive_rating_NAT) did not conform to a normal distribution (p < 0.05) (Table 4). Therefore, in 3 pairs (Points_per_possession, Defensive_rebounds, Offensive_rating), the non-parametric Wilcoxon Signed Ranks Test was applied, while in the remaining 15 Paired Samples t-Test was employed.

Tables 5 and 6 indicate that, for all 18 variables contributing significantly to the team's wins or losses, there is a statistically significant difference between the values in the Euroleague and the corresponding values in the domestic leagues.

From table 7 it can be seen that the differences between the teams regarding their performance in the Euroleague and the domestic leagues were: a) moderate in the variables threept_field_goals_percent (d=0.578) andSteals_to_turnovers (d=0.751), b) large in the variables Field_goals_percent (d= 1.151), Effective_field_goal_percentage (d=1.122), Contested_field_goals_made (d=0.866), Contested_field_goals_percent (d=1.149) and Points_off_turnovers (d=0.908), and c) very large in the variables Points (d=2.14), Points_per_possession (d=1.578 ), Rebounds (d=1.792), Defensive_rebounds (d=1.837), Offensive_rating (d=1.803), Defensive_rating (d=-1.408), Defensive_Efficiency (d=-0.154), Net_rating (d=1.943), Assists_to_turnovers (d= 1.341) and True_shooting_percentage (d=1.468).

Discussion

Through a robust application of advanced machine learning techniques, including RF, LR, and SVMs, complemented by the Boruta FS and SHAP model for interpretability, this research provides a sophisticated understanding of which factors most strongly predict match outcomes and how these factors differ by competition context. Furthermore, the findings of this research significantly contribute to the sports analytics domain by elucidating how KPIs differ between the Euroleague and national leagues. This dual analysis not only bridges a critical bibliographic gap but also enhances our comprehension of how strategic and performance demands shift depending on the competitive environment.

The effectiveness of field goals, as reflected by effective field goal percentage (eFG%), stood out as a critical factor in both competition settings. Literature corroborates this finding, indicating that eFG% is often a determinant of game outcomes across various basketball leagues (Li et al., 2023; Malarranha et al., 2013; Mandić et al., 2019). This suggests a universal applicability of this metric across competitive levels, reinforcing its utility in strategic game planning and player performance evaluation. However, while the importance of eFG% is welldocumented, our analysis further enriches this understanding by quantifying its variable impact between different leagues, potentially reflecting variations in defensive strategies or offensive efficiency that are unique to each league's style of play.

Furthermore, assists are a key metric in basketball, often serving as an indicator of team cohesion and effectiveness in offensive execution. The significant finding in our study for assists-to-turnovers, points to the critical role of efficient ball handling and distribution in securing game victories. This observation resonates with the insights from Terner and Franks (2021), who discuss the predictive power of assist rates in assessing team performance. By enhancing our understanding of the strategic deployment of assists within game play, particularly in creating scoring opportunities while minimizing turnovers, teams can better orchestrate their offensive plays and improve overall team dynamics.

Rebounds have consistently been identified as a critical factor in basketball analytics, often correlating strongly with game outcomes due to their direct impact on possession control. The significant emphasis on rebounds reflects findings similar to those in Han and Choi (2020) and Leicht et al. (2017), where rebounding prowess, particularly defensive rebounds, was associated with winning outcomes. This aligns with our study's findings where rebounds, particularly defensive rebounds, showed a notable difference between winning and losing teams. This suggests that successful teams possess a robust rebounding strategy, reinforcing the need for teams to focus on enhancing this aspect of their game to increase their chances of victory.

Defensive metrics, particularly defensive ratings, also highlight a distinct influence in the Euroleague compared to domestic leagues. This divergence points to the tactical sophistication and higher stakes associated with Euroleague games, where minimizing opponent scores is paramount. The significant role of defensive strategies in the Euroleague aligns with broader findings by Milanović et al. (2019), who point out the critical nature of defensive actions in influencing game outcomes at the Olympic level. This divergence in defensive impact between leagues may also reflect differences in player skills, coaching strategies, or even the physicality of the game, which varies widely across different competitive environments (Muro et al., 2020). Furthermore, steals, indicative of aggressive and effective defensive play, were also highlighted as a significant factor in our study, aligning with insights from Milanović et al. (2019), who emphasized the importance of steals in contributing to winning performances at the Olympic level.

These nuanced understandings of KPIs in basketball underscore the adaptability and strategic foresight required in coaching and game management. They also suggest that while certain performance metrics like eFG% and rebounds are universally important, their specific applications and impact can vary markedly between different types of competitions. Such insights are invaluable for refining training programs and game-day strategies, ensuring that teams are better equipped to compete at their peak in varied competitive environments. Collectively, the study enriches the strategic toolkit available to basketball teams, offering nuanced insights that can drive more informed decision-making in both preparation and real-time game management.

Despite the comprehensive analysis and significant contributions of this study, several limitations must be acknowledged. Firstly, the analysis is confined to the performance data from the Euroleague and National leagues over specific seasons, which may not fully encapsulate the dynamics or potential changes in playing styles and team strategies that evolve over time (Barreira & Morgado, 2023; García-Izquierdo et al., 2012). Additionally, the study relies on the accuracy and completeness of the data provided by the Instat Basketball platform. While this platform is highly regarded (Bustamante-Sánchez et al., 2022), the potential for missing or inaccurately recorded data could affect the validity of the findings. Furthermore, the machine learning models employed, although sophisticated, are dependent on the assumption that past performances are indicative of future outcomes, which in sports can be a precarious assumption due to the unpredictable nature of games and player performances. Lastly, this study employs a static analytical method that, while commonly used, provides only a snapshot of performance, disregarding the dynamic nature of matches (Pratas et al., 2018). In contrast, dynamic analysis considers the game's state at each moment, providing a more comprehensive understanding of performance patterns and outcomes (Prieto et al., 2015). These limitations suggest areas for further research, including longitudinal studies and the use of the dynamic method to enrich the understanding of performance differences in basketball.

Conclusions

This study represents a significant advancement in the field of sports analytics by systematically analyzing and comparing KPIs in professional basketball across two distinct competitive environments: the Euroleague and national leagues. Through a robust application of advanced machine learning techniques, this research has not only elucidated the most influential KPIs but also highlighted the variation in their impact between different competitions. The empirical validation of these findings, supported by sophisticated models such as RF, LR, and SVMs, ensures that our conclusions are both theoretically sound and practically viable, underpinning the study's substantial contribution to the literature.

The importance of this research to the specialized scientific community cannot be underestimated. It fills a critical bibliographic gap by providing a nuanced understanding of how PIs vary significantly between international and domestic competitions. This differentiation is crucial for developing more tailored training programs and game strategies that are sensitive to the specific demands of each competition type. Moreover, the use of SHAP values for interpretability allows for a deeper insight into the data, helping researchers and practitioners alike understand the complex dynamics that influence game outcomes.

For coaches, analysts, and scouters, the practical applications of this study are manifold. By identifying which performance metrics are most indicative of success in different league settings, coaching staff can optimize their strategies to exploit these insights effectively. For instance, the emphasis on effective field goal percentage (eFG%) and defensive ratings in the Euroleague suggests a need for focused training on shooting efficiency and defensive tactics in preparation for these high-caliber matches. Additionally, the impact of assists and rebounds provides strategic data points for adjusting team dynamics and player roles depending on the opponent's league standard.

In conclusion, this research not only advances our understanding of sports analytics by integrating complex ML tools into the analysis of basketball performance but also offers actionable insights that can significantly enhance the preparation and execution of basketball strategies across various levels of professional play. The findings from this study serve as a cornerstone for future explorations into the adaptive nature of sports performance analysis, paving the way for more sophisticated and context-sensitive approaches in sports science.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest.

Sidebar

Published online: June 30, 2024

Accepted for publication : June 15, 2024

Corresponding Author: SPYRIDON PLAKIAS, E-mail: [email protected]

References

Barreira, J., & Morgado, G. (2023). Home advantage in basketball: A longitudinal analysis of the NBA playoffs (1946-2022). Retos: nuevas tendencias en educación física, deporte y recreación(47), 691-694. https://dialnet.unirioja.es/servlet/articulo?codigo=8707075

Bourdas, D. I., Mitrousis, I., Zacharakis, E. D., & Travlos, A. K. (2022). Home-audience advantage in basketball: evidence from a natural experiment in Euro League games during the 2019-2021 Covid-19 era. Journal of Physical Education and Sport, 22(7), 1761-1771. https://doi.org/10.7752/jpes.2022.07220

Bustamante-Sánchez, Á., Gómez, M. A., & Jiménez-Saiz, S. L. (2022). Game location effect in the NBA: A comparative analysis of playing at home, away and in a neutral court during the COVID-19 season. International Journal of Performance Analysis in Sport, 22(3), 370-381. https://doi.org/10.1080/24748668.2022.2062178

Butterworth, A., O'Donoghue, P., & Cropley, B. (2013). Performance profiling in sports coaching: a review. International Journal of Performance Analysis in Sport, 13(3), 572-593. https://doi.org/10.1080/24748668.2013.11868672

Ermiş, E., Ermiş, A., Erilli, N. A., & Konca, E. (2019). The association between basketball players'times in the game and their performance: A comparison of Euroleague-Eurobasket. Spor ve Performans Araştırmaları Dergisi, 10(2), 114-122. https://doi.org/10.17155/omuspd.516270

García-Izquierdo, A. L., Ramos-Villagrasa, P. J., & Navarro, J. (2012). Dynamic criteria: A longitudinal analysis of professional basketball players' outcomes. The Spanish journal of psychology, 15(3), 1133-1146. https://doi.org/10.5209/rev_SJOP.2012.v15.n3.39403

Han, D., & Choi, H. (2020). The identification of optimal data range for the discrimination between won and lost. Journal of the Korea Computer Information Society, 25(7), 103-111. https://doi.org/https://www.dbpia.co.kr/Journal/articleDetail?nodeId=NODE09409721

Hughes, M. D., & Bartlett, R. M. (2002). The use of performance indicators in performance analysis. Journal of sports sciences, 20(10), 739-754. https://doi.org/10.1080/026404102320675602

Izzo, R., Varde'i, C. H., Righi, E., Cejudo, A., & Giovannelli, M. (2023). Statistical implications of the pick technique in a basketball match performance analysis. Journal of Physical Education and Sport, 23(3), 569-578. https://doi.org/10.7752/jpes.2023.03071

Kapsalis, M., Plakias, S., Kyranoudis, A., Zarkadoula, A., Lathoura, A., & Tsatalas, T. (2023). Exploring the impact of possession-based performance indicators on goal scoring in elite football leagues. Journal of Physical Education and Sport, 23(8), pp. 2004 - 2015. https://doi.org/10.7752/jpes.2023.08231

Lampis, T., Ioannis, N., Vasilios, V., & Stavrianna, D. (2023). Predictions of european basketball match results with machine learning algorithms. Journal of Sports Analytics(Preprint), 1-20. https://doi.org/10.3233/JSA-220639

Leicht, A. S., Gomez, M. A., & Woods, C. T. (2017). Team performance indicators explain outcome during women's basketball matches at the Olympic Games. Sports, 5(4), 96. https://doi.org/10.3390/sports5040096

Li, M.-z., Lu, Y., Chen, Z.-h., Yang, Q., & Tao, Y.-l. (2023). Analysis of Key Performance Indicators and Major Winning Rules in Chinese Basketball Professional League. https://doi.org/10.21203/rs.3.rs-3407093/v1

Malarranha, J., Figueira, B., Leite, N., & Sampaio, J. (2013). Dynamic modeling of performance in basketball. International Journal of Performance Analysis in Sport, 13(2), 377-387. https://doi.org/10.1080/24748668.2013.11868655

Mandić, R., Jakovljević, S., Erčulj, F., & Štrumbelj, E. (2019). Trends in NBA and Euroleague basketball: Analysis and comparison of statistical data from 2000 to 2017. PloS one, 14(10), e0223524. https://doi.org/10.1371/journal.pone.0223524

Mikołajec, K., Banyś, D., Żurowska-Cegielska, J., Zawartka, M., & Gryko, K. (2021). How to win the basketball EuroLeague? Game performance determining sports results during 2003–2016 matches. Journal of Human kinetics, 77(1), 287-296. https://doi.org/10.2478/hukin-2021-0050

Milanović, D., Uzelac, N., & Šalaj, S. (2019). Game efficiency indicators of Olympic basketball performance. Acta kinesiologica, 13(1), 17-21. https://akinesiologica.com/game-efficiency-indicators-of-olympicbasketball-performance/

Moustakidis, S., Plakias, S., Kokkotis, C., Tsatalas, T., & Tsaopoulos, D. (2023). Predicting Football Team Performance with Explainable AI: Leveraging SHAP to Identify Key Team-Level Performance Metrics. Future Internet, 15(5), 174. https://doi.org/10.3390/fi15050174

Muro, I., Dilicikik, U., & Kocaoglu, B. (2020). Medical Coverage of Basketball Events: From Local Competitions to European, World Championships and Olympic Games. Basketball Sports Medicine and Science, 103-110. https://doi.org/10.1007/978-3-662-61070-1_9

Oliveira-Da-Silva, L., Sedano-Campo, S., & Redondo-Castán, J. C. (2013). Características del esfuerzo en competición en jugadoras de baloncesto de élite durante las fases finales de la Euroliga y el Campeonato del Mundo.[The competitive demands of elite female basketball during the play-offs of the Euroleague and World Championship]. RICYDE. Revista Internacional de Ciencias del Deporte., 9(34), 360-376. https://doi.org/10.1136/bjsm.2006.032318

Paulauskas, R., Masiulis, N., Vaquera, A., Figueira, B., & Sampaio, J. (2018). Basketball game-related statistics that discriminate between European players competing in the NBA and in the Euroleague. Journal of Human kinetics, 65(1), 225-233. https://doi.org/10.2478/hukin-2018-0030

Plakias, S., Moustakidis, E., Mitrotasios, M., Kokkotis, C., Tsatalas, T., Papalexi, M., Giakas, G., & Tsaopoulos, D. (2023a). A Multivariate and cluster analysis of diverse playing styles across European Football Leagues. Journal of Physical Education and Sport, 23(7), pp. 1631-1641. https://doi.org/10.7752/jpes.2023.07200

Plakias, S., Moustakidis, S., Mitrotasios, M., Kokkotis, C., Tsatalas, T., Papalexi, M., Giakas, G., & Tsaopoulos, D. (2023b). Analysis of playing styles in European football: insights from a visual mapping approach. Journal of Physical Education and Sport, 23(6), pp. 1385 - 1393. https://doi.org/10.7752/jpes.2023.06169

Plakias, S., Tsatalas, T., Armatas, V., Tsaopoulos, D., & Giakas, G. (2024). Tactical Situations and Playing Styles as Key Performance Indicators in Soccer. Journal of Functional Morphology and Kinesiology, 9(2), 88. https://doi.org/10.3390/jfmk9020088

Pratas, J. M., Volossovitch, A., & Carita, A. I. (2018). Goal scoring in elite male football: A systematic review. Journal of Human Sport and Exercise 13(1), 218-230. https://doi.org/10.14198/jhse.2018.131.19

Prieto, J., Gómez, M.-Á., & Sampaio, J. (2015). From a static to a dynamic perspective in handball match analysis: A systematic review. The Open Sports Sciences Journal, 8(1). https://doi.org/10.2174/1875399X01508010025

Terner, Z., & Franks, A. (2021). Modeling player and team performance in basketball. Annual Review of Statistics and Its Application, 8, 1-23. https://doi.org/10.1146/annurev-statistics-040720-015536

Tormo, J. V. G., Manzano, D. P., Jiménez, A. V., & Rábago, J. C. M. (2015). INCIDENCIA DE LOS TIROS LIBRES EN PARTIDOS DE BALONCESTO PROFESIONAL [Incidence of free throws in professional basketball games]. E-Balonmano. com: Revista de Ciencias del Deporte, 11(1), 73-82. http://hdl.handle.net/10662/7444

Zhou, W., Sansone, P., Jia, Z., Gomez, M.-A., & Li, F. (2024). Determining the key performance indicators on game outcomes in NBA based on quantile regression analysis. International Journal of Performance Analysis in Sport, 1-16. https://doi.org/10.1080/24748668.2024.2325846

Appendices Appendix A:

Appendix B: