Football Analytics: Assessing the Correlation between Workload, Injury and Performance of Football Players in the English Premier League
Abstract
:1. Introduction
1.1. Motivation
1.2. Background and Context
1.3. Aims and Objectives
- What is the correlation between payer workload, player characteristics, and performance of football players in the English Premier League?
- How do game/match-related variables impact football players’ performance in the English Premier League, and what are the essential variables contributing towards the prediction of player performance?
- How do game/match-related variables impact the occurrence of injuries among football players in the English Premier League, and what are the important variables contributing to injury prediction?
- How do various factors influence injury occurrence among players?
- Correlation Exploration: Conduct detailed statistical analyses, such as descriptive analytics and correlation analysis, to identify and quantify the complex correlations between player workload, individual traits, and performance measures.
- Predictive Modelling: Develop and integrate advanced ML models to assess the prediction capacity of diverse variables. Determine the essential characteristics that have a major impact on the forecast of player performance (goals and assists), providing insights into the drivers of football success.
- Injury Occurrence Analysis: Examine the impact of game and match-related variables on the occurrence of injuries among English Premier League football players. Use rigorous statistical tools to determine the key variables contributing to injury incidence.
2. Literature Review
2.1. Introduction
2.2. Research and Findings
- Statistical Metrics: Analytics involving the collection of statistical data such as goals scored, goals assisted, passes completed, etc., which reflects a player’s overall contribution to the team’s performance.
- Position-Specific Analysis: Analytics estimating the effectiveness of various players across different areas of the playing field, allowing them to assess their strong and weak areas, which can highlight potential improvements.
- Physical performance: Data related to player physical attributes, such as sprint speed, distance covered, and high-intensity runs, help gauge a player’s fitness level and work rate during matches.
- Video Analysis: In addition to statistical data, video analysis is used to evaluate a player’s decision-making, movement, positioning, and technical skills during matches.
- Injury Prevention: Understanding the relationship between workload and injury is crucial for developing effective injury prevention strategies. By identifying workload thresholds and patterns associated with higher injury risks, clubs and medical staff can implement targeted measures to reduce the likelihood of injuries.
- Performance Optimization: The correlation between workload and performance is a critical aspect of player management. Balancing the right level of workload can positively impact player performance, ensuring optimal physical and technical abilities on the pitch.
- Player Management: A deep understanding of the workload-injury-performance relationship allows for better player management.
3. Methodology
3.1. Objective
3.2. Dataset
- Personal Information: Player Name, Club Name, Age, Position etc.
- Individual Workload Features: Minutes Played, Matches Played, Matches Started, and 90 Minutes Played.
- Individual Performance Features: Goals Scored, Assists, Goals Plus Assists etc.
- Individual Injury Occurrences: Injured (Injured or Not Injured), Injury Reason, Injury Occurrences etc.
3.3. Methods and Structure
3.4. Data Preprocessing
- (a)
- Irrelevant variables: The numerical data collected from the sources include several variables irrelevant to the analysis, such as expected goals (the number of goals a player is likely to score in the current season by analyzing his previous season’s performance), nationality, and the number of yellow and red cards acquired by player throughout the season. These variables need to be removed before conducting the analysis. The process of feature selection in this study was guided by a combination of domain expertise and statistical analysis to ensure that only the most relevant features were included in our models. Initially, features were chosen based on their known impact on player performance and injury risk from existing sports science literature. This foundational selection was further refined through exploratory data analysis, where correlation matrices and preliminary regression models helped identify features with significant predictive power and minimal collinearity.
- (b)
- Minutes played: The data include players who played as little as one minute of a match in the season. This will result in the data analysis being irrelevant and misleading. To address this, we set a parameter to include only those data that qualify for data analysis. We set the parameter to a minimum of 1140 min (setting an average of 30 min for each match played, so 38 matches imply a total of 38 multiplied by 30, which is 1140 min). This minimum threshold was chosen because it represents a substantial portion of the match time to contribute meaningfully to team dynamics while allowing for the inclusion of players who may not start every match but are regular contributors. This method not only reduces the risk of analyzing data that might not truly reflect a player’s impact but also helps to eliminate incomplete records from the dataset.Minutes played = Total number of matches in the season × Minimum standard set = 38 × 30 = 1140 min
- (c)
- Player position: Due to the different roles that midfielders and forwards play on the field, the research will focus on these two groups of players when predicting performance. The central planners of attacking moves are the midfielders (MF) and forwards (FW) players, who create opportunities for goals through assists and their own goals. They are the perfect subjects for this research because their performance measures are, by nature, focused on attack and creativity. Unlike goalkeepers or defenders, whose contributions are assessed differently, analyzing MF and FW players’ offensive prowess offers deeper insights into the fundamental elements of football performance that directly impact a team’s ability to win. This tactical decision enables us to focus on the players whose actions impact goal-scoring and offensive playmaking most.
3.4.1. Feature Engineering
- (a)
- Player Workload:The ‘player workload’ variable classifies football players as ‘Rare Starter’, ‘Average Starter’, and ‘Frequent Starter’, depending on how frequently they start matches. This variable provides insights into workload differences and their consequences on injuries and performance. These cut-off points were derived from a review of historical data and existing literature on player health and performance, providing a structured approach to monitor and manage player workload effectively.Number of matches started:
- Less than 10 ⇒ Rare starter
- More than 10 but less than or equal to 25 ⇒ Average starter
- More than 25 ⇒ Frequent starter
- (b)
- Player Usage:Based on the number of minutes played, the ‘Player Usage’ variable classifies football players as ‘Squad Rotation Players’, ‘Sporadic Players’, or ‘Crucial Players’. This emphasizes the importance of specific players to their teams and provides information about their usage’s impact on performance, fitness, and injury risk.Number of minutes played:
- Less than 1140 min ⇒ Sporadic Player
- Between 1140 and 2280 (inclusively) minutes ⇒ Squad Rotation Player
- More than 2280 min ⇒ Crucial Player
- (c)
- Age Category:Based on age, the ‘Age Category’ variable classifies football players as ‘Youngster’, ‘Prime’, or ‘Veteran’. This sheds light on the impact of age on career duration, injury risk, and performance. It aids players in understanding how age and experience affect their roles and output.Age falling between:
- 16 and 23⇒ Youngster
- 24 and 31 ⇒ Prime
- 32 and above ⇒ Veteran
- (d)
- Average Minutes per MatchBy determining a player’s average minutes played per match, the ‘Average Minutes per Match’ variable offers a fair assessment of player playing time. This helps in determining the consistency of player participation in games, which may affect injury risk and performance. It serves as a starting point when examining how player workload, injury, and performance are related.Average Min/Match = (Minutes Played)/(Matches Played)
3.4.2. Dummy Variables
3.4.3. Train-Test Strategy
3.4.4. Sampling Methods
3.5. Descriptive Statistics
3.6. Correlation Matrix
3.7. Machine Learning (ML)
3.7.1. Naive Model
3.7.2. Decision Tree
3.7.3. Random Forests
3.7.4. K-Nearest Neighbors (KNN)
3.7.5. Gradient Boosting
3.7.6. Ridge Regression
3.7.7. XGBoost
3.8. Hyperparameter Tuning
3.9. Evaluation Metric
- In the case of numerical ‘Goals and Assists’, RMSE is an ideal choice since it estimates the average size of forecast errors. It gives a comprehensive assessment of how well the ML model’s predictions match the actual numerical outcomes. Because RMSE prioritizes avoiding both overestimation and underestimation, it is appropriate for evaluating the prediction accuracy of models attempting to estimate continuous variables such as goals and assists.
- Accuracy is a suitable choice for categorical ‘Injured’ and ‘Not Injured’ outcomes since it represents the proportion of accurately predicated cases. This metric is critical when it comes to appropriately classifying the occurrence or non-occurrence of an event. Given the significance of accurately detecting injuries, accuracy clearly indicates the model’s ability to categorize these occurrences.
4. Results
4.1. Correlation Analysis
4.2. Predicting Goals and Assists
4.2.1. Model Performance Assessment
Residual Analysis
Training and Testing Assessment
- Decision Tree versus Random Forests: The Random Forest Regressor demonstrates superior initial performance, evidenced by its lower RMSE in both the training and testing sets, suggesting that it is a more effective model straight away when compared to the Decision Tree Regressor. Its smaller discrepancy between training and cross-validation RMSEs signals a reduced tendency toward overfitting, an advantage over the Decision Tree. For both models, adding more data improves the model’s performance, but the rate of improvement slows down, which is typical as a model starts to reach its performance limit with the given features and model complexity. Notably, the Random Forest Regressor shows less variability in its testing RMSE, which is illustrated by a narrower confidence interval, indicating a more consistent performance regardless of the training set it encountered.
- Gradient Boosting versus Ridge Regression: The Gradient Boosting model has a considerable and continuous gap between training and testing RMSE. The training RMSE drops significantly at first, demonstrating the model’s ability to fit the training data effectively. However, the testing RMSE improves more slowly, constantly remaining higher than the training RMSE. This difference indicates overfitting, as the model struggles to generalize to new data. The Ridge Regression learning curve, on the other hand, begins with a significant gap between training and testing RMSE. The gap narrows dramatically as training progresses, and the two RMSE curves converge. The training RMSE gradually rises while the testing RMSE falls, indicating better generalization. Ridge Regression finds a better balance between fitting the training data and generalizing to new data based on this behavior. While the RMSE values are not the lowest among the models, the model’s ability to resist overfitting is a significant benefit.
4.2.2. Model Performance Comparison Based on RMSE
4.2.3. Feature Importance
4.3. Predicting Injuries
4.3.1. Model Evaluation
- Injury measurements are not relevant for the Naive Model because it uses goals/assists prediction as a baseline exclusively.
- Since Gradient Boosting and Ridge Regression are suitable for managing continuous outcomes, they were only applied to objectives and assistance prediction. For these models, injury measurements are therefore irrelevant.
- As XGBoost and K-Nearest Neighbors are appropriate at classification tasks, they were especially used for injury prediction. Consequently, RMSE values for objectives and aids are not relevant.
- Decision-Tree and Random Forest can be comprehensive models, but they may sometimes underperform in any of metrics.
4.3.2. Feature Importance
5. Discussion and Findings
5.1. Predictors for Player Performance (Goals and Assists)
5.1.1. Matches Played
- Strong Positive Correlations with Workload Metrics: It became clear that player performance metrics, particularly goals and assists, exhibit strong positive correlations with workload metrics, including elements like minutes played, matches played, and 90s (minutes played divided by 90). This result is consistent with common sense because more time spent on the field naturally gives players more chances to assist their teammates and score goals.
- Matches Played (MP) with emphasis: Within the subgroup of match load measures, matches played (MP) stood out as especially significant. Compared to other measures like minutes played (Min) and 90s (minutes played per match), which had correlations of 0.23, MP had a higher correlation coefficient of 0.34. This suggests that a player’s ability to score goals and provide assists is most significantly influenced by the sheer number of games in which they take part, regardless of how much time they spend on the field or whether they are a starter.
5.1.2. Squad
- The prominence of Top Teams: Players from the season’s top four teams (Manchester City, Manchester United, Liverpool, and Leicester City) are marked in bold. This distinction is critical because it emphasizes that being a part of a high-performing team has a major impact on a player’s ability to accumulate goals and assists. In this case, these four teams account for half of the top 20 performers (10 out of 20, as emphasized in bold in Table 8), demonstrating their dominance in player performance metrics.
- Variety of Positions: The table includes players from numerous positions, with forwards contributing the most goals and assists. Midfielders like Kevin De Bruyne, Bruno Fernandez, and Jack Harrison are also prominently featured in the top 20, highlighting their versatile responsibilities in both scoring and creating goals.
- Individual Brilliance: The list features some of the league’s most prolific goal scorers and playmakers, including Harry Kane and Bruno Fernandes, at the top. These players are recognized for their extraordinary abilities and consistent impact on matches, routinely scoring goals and making assists.
- Balance and Competitiveness: The presence of players such as Patrick Bamford and Ollie Watkins, who play for clubs other than the conventional top tier, demonstrates that the Premier League retains a competitive and diverse player environment. This type of balance brings interest to the league by allowing developing talent to flourish.
5.1.3. Age Category
5.2. Predicting Injuries
5.2.1. Squad
5.2.2. Average Minutes per Match
5.2.3. Age
6. Conclusions
6.1. Key Findings
6.2. Contribution to the Field of Sports Science and Analytical Research
6.3. Impact on Football Clubs
6.4. Applicability to Other Domains
6.5. Limitations
6.6. Recommendations for Future Work
Author Contributions
Funding
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
References
- Clemente, F.M.; Martins, F.M.L.; Mendes, R.S.; Figueiredo, A.J. A systemic overview of football game: The principles behind the game. J. Hum. Sport Exerc. 2015, 9, 656–667. [Google Scholar] [CrossRef]
- Asif, R.; Zaheer, M.T.; Haque, S.I.; Hasan, M.A. Football (soccer) analytics: A case study on the availability and limitations of data for football analytics research. Int. J. Comput. Sci. Inf. Secur. 2016, 14, 516. [Google Scholar]
- Chazan-Pantzalis, V.; Tjortjis, C. Sports Analytics for Football League Table and Player Performance Prediction. In Proceedings of the 2020 11th International Conference on Information, Intelligence, Systems and Applications, Piraeus, Greece, 15–17 July 2020; pp. 1–8. [Google Scholar]
- Rodrigues, F.; Pinto, Â. Prediction of football match results with Machine Learning. Procedia Comput. Sci. 2022, 204, 463–470. [Google Scholar] [CrossRef]
- Seidenschwarz, P.; Rumo, M.; Probst, L.; Schuldt, H. A Flexible Approach to Football Analytics: Assessment, Modeling and Implementation. In Proceedings of the 12th International Symposium on Computer Science in Sport (IACSS 2019); Springer International Publishing: Cham, Switzerland, 2020. [Google Scholar]
- Windt, J.; Gabbett, T.J. How do training and competition workloads relate to injury? The workload-injury aetiology model. Br. J. Sports Med. 2017, 51, 428–435. [Google Scholar] [CrossRef] [PubMed]
- Cefis, M.; Carpita, M. Football Analytics: Performance analysis differentiate by role. In Third International Conference on Data Science & Social Research Book of Abstracts; CIRPAS and University of Bari Aldo Moro: Bari, Italy, 2020; p. 22. [Google Scholar]
- Javed, D.; Jhanjhi, N.Z.; Khan, N.A. Football Analytics for Goal Prediction to Assess Player Performance. In Proceedings of Innovation and Technology in Sports; Springer Nature: Singapore, 2023; pp. 245–257. [Google Scholar]
- Mead, J.; O’Hare, A.; McMenemy, P. Expected goals in football: Improving model performance and demonstrating value. PLoS ONE 2023, 18, e0282295. [Google Scholar] [CrossRef]
- Baboota, R.; Kaur, H. Predictive analysis and modelling football results using machine learning approach for English Premier League. Int. J. Forecast. 2019, 35, 741–755. [Google Scholar] [CrossRef]
- Gronwald, T.; Klein, C.; Hoenig, T.; Pietzonka, M.; Bloch, H.; Edouard, P.; Hollander, K. Hamstring injury patterns in professional male football (soccer): A systematic video analysis of 52 cases. Br. J. Sports Med. 2021, 56, 165–171. [Google Scholar] [CrossRef] [PubMed]
- Howle, K.; Waterson, A.; Duffield, R. Injury Incidence and Workloads during congested Schedules in Football. Int. J. Sports Med. 2019, 41, 75–81. [Google Scholar] [CrossRef] [PubMed]
- Sarlis, V.; Tjortjis, C. Sports Analytics: Data Mining to Uncover NBA Player Position, Age, and Injury Impact on Performance and Economics. Information 2024, 15, 242. [Google Scholar] [CrossRef]
- Alayón, S.; Hernández, J.; Fumero, F.J.; Sigut, J.F.; Díaz-Alemán, T. Comparison of the Performance of Convolutional Neural Networks and Vision Transformer-Based Systems for Automated Glaucoma Detection with Eye Fundus Images. Appl. Sci. 2023, 13, 12722. [Google Scholar] [CrossRef]
- Xu, M.; Watanachaturaporn, P.; Varshney, P.K.; Arora, M.K. Decision tree regression for soft classification of remote sensing data. Remote Sens. Environ. 2005, 97, 322–336. [Google Scholar] [CrossRef]
- Liaw, A.; Wiener, M. Classification and Regression by Randomforest. R News 2002, 2, 18–22. [Google Scholar]
- Guo, G.; Wang, H.; Bell, D.; Bi, Y.; Greer, K. KNN Model-Based Approach in Classification; Springer: Cham, Switzerland, 2003; pp. 986–996. [Google Scholar]
- Friedman, J.H. Stochastic gradient boosting. Comput. Stat. Data Anal. 2002, 38, 367–378. [Google Scholar] [CrossRef]
- de Vlaming, R.; Groenen, P.J. The Current and Future Use of Ridge Regression for Prediction in Quantitative Genetics. Biomed. Res. Int. 2015, 2015, 143712. [Google Scholar] [CrossRef] [PubMed]
- Abdurrahman, G.; Sintawati, M. Implementation of xgboost for classification of parkinson’s disease. J. Phys. Conf. Ser. 2020, 1538, 012024. [Google Scholar] [CrossRef]
- Belete, D.M.; Huchaiah, M.D. Grid search in hyperparameter optimization of machine learning models for prediction of HIV/AIDS test results. Int. J. Comput. Appl. 2022, 44, 875–886. [Google Scholar] [CrossRef]
- McKeown, G. To Build a Top Performing Team, Ask for 85% Effort. Available online: https://hbr.org/2023/06/to-build-a-top-performing-team-ask-for-85-effort (accessed on 14 August 2024).
Study Title | Author(s) | Category | Aim | Year |
---|---|---|---|---|
How do training and competition workloads relate to injury? | Windt and Gabbett [6] | Workload, Injuries and Performance in Football | Present a new framework for managing football injuries | 2017 |
Injury Incidence and Workloads during congested Schedules in Football | Howle et al. [12] | Workload, Injuries and Performance in Football | Examine the relationship between injury incidence and workloads in congested football schedules | 2019 |
Predictive analysis and modelling football results using machine learning approach for English Premier League | Baboota & Kaur [10] | Performance Analysis in Football | Une machine learning for predictive analysis of football match outcomes | 2019 |
Football Analytics: Performance analysis differentiate by role | Cefis and Carpita [7] | Performance Analysis in Football | Develop composite indices for performance assessment in football | 2020 |
Hamstring injury patterns in professional male football | Gronwald et al. [11] | Workload, Injuries, and Performance in Football | Identify factors contributing to acute hamstring injuries in professional male football players | 2021 |
Expected goals in football: Improving model performance and demonstrating value | Mead et al. [9] | Performance Analysis in Football | Improve Expected Goals (xG) modeling for assessing team success in football | 2023 |
Football Analytics for Goal Prediction to Assess Player Performance | Javed et al. [8] | Performance Analysis in Football | Greate a Goal Prediction Model and assess player performance in football | 2023 |
Model | Hyper Parameter | RMSE |
---|---|---|
Naive Model | 6.62 | |
Decision Tree (DC_1) | Default | 4.77 |
Decision Tree (DC_2) | Max depth: None Min samples leaf: 4 Min samples split: 2 | 4.41 |
Random Forest (RF_1) | Default | 4.33 |
Random Forest (RF_2) | Max depth: None Min samples leaf: 2 Min samples split: 5 N estimators: 50 | 4.24 |
Gradient Boosting (GB_1) | Default | 4.13 |
Gradient Boosting (GB_2) | Learning rate: 0.1 Max depth: 3 Min samples leaf: 4 Min samples split: 10 N estimators: 50 | 4.05 |
Ridge Regression (RR_1) | Default | 3.91 |
Ridge Regression (RR_2) | alpha: 10 | 3.90 |
Feature Importance | Decision Tree | Random Forest | Gradient Boosting |
---|---|---|---|
Minutes Played (Min) | 0.024 | 0.397 | 0.305 |
Matches Played (MP) | 0.022 | 0.047 | 0.054 |
Starts | 0.073 | 0.412 | 0.058 |
90 Minutes Played (90s) | 0.874 | 0.114 | 0.552 |
Age | 0.006 | 0.030 | 0.032 |
Model | Accuracy | Precision | Recall | AUC Score |
---|---|---|---|---|
Decision Tree (DTC_2) | 0.72 | 0.58 | 0.35 | 0.66 |
Random Forest (RFC_2) | 0.69 | 0.50 | 0.50 | 0.70 |
XGBoost (XGB_2) | 0.72 | 0.55 | 0.60 | 0.65 |
K-Nearest Neighbors (KNN_2) | 0.63 | 0.405 | 0.50 | 0.54 |
Model (Baseline) | RMSE (Goals/Assists) | Accuracy (Injury) | Precision (Injury) | Recall (Injury) | AUC (Injury) |
---|---|---|---|---|---|
Naive Model | 6.62 | NA | NA | NA | NA |
Gradient Boosting | 4.05 | NA | NA | NA | NA |
Ridge Regression | 3.90 | NA | NA | NA | NA |
Model (Injury investigations) | RMSE (Goals/Assists) | Accuracy (Injury) | Precision (Injury) | Recall (Injury) | AUC (Injury) |
XGBoost | NA | 0.72 | 0.55 | 0.60 | 0.65 |
K-Nearest Neighbors | NA | 0.63 | 0.405 | 0.50 | 0.54 |
Model (Comprehensive) | RMSE (Goals/Assists) | Accuracy (Injury) | Precision (Injury) | Recall (Injury) | AUC (Injury) |
Decision Tree | 4.41 | 0.72 | 0.58 | 0.35 | 0.66 |
Random Forest | 4.24 | 0.69 | 0.50 | 0.50 | 0.70 |
Feature | Random Forest (RFC_2) | Decision Tree (DTC_2) | XGBoost (XGB_2) |
---|---|---|---|
Position | 0.06 | 0.11 | 0.03 |
Squad | 0.16 | 0.26 | 0.27 |
Age | 0.11 | 0.12 | 0.20 |
Matches Played | 0.12 | 0.08 | 0.12 |
Starts | 0.10 | 0.04 | 0.11 |
Minutes Played | 0.12 | 0.04 | 0.08 |
Average Minutes/Match | 0.12 | 0.25 | 0.11 |
90s | 0.16 | 0.09 | 0.07 |
Player Workload | 0.00 | 0.00 | 0.00 |
Player Usage | 0.01 | 0.00 | 0.01 |
Age Category | 0.03 | 0.00 | 0.01 |
Player | 2020/21 | 2021/22 | 2022/23 | |||
---|---|---|---|---|---|---|
Matches Played | Gls_Ast | Matches Played | Gls_Ast | Matches Played | Gbs_Ast | |
Harry Kane | 35 | 37 | 37 | 26 | 38 | 33 |
Mohammed Salah | 37 | 27 | 35 | 36 | 38 | 31 |
Bruno Fernandes | 37 | 30 | 36 | 16 | 37 | 16 |
Player | Goals + Assists | Team | Position |
---|---|---|---|
Harry Kane | 37 | Tottenham | Forward |
Bruno Fernandes | 30 | Manchester United | Midfielder |
Son Heung-min | 27 | Tottenham | Forward |
Mohamed Salah | 27 | Liverpool | Forward |
Jamie Vardy | 24 | Leicester City | Forward |
Patrick Bamford | 24 | Leeds United | Forward |
Marcus Rashford | 20 | Manchester United | Forward |
Ollie Watkins | 19 | Aston Villa | Forward |
Kevin De Bruyne | 18 | Manchester City | Midfielder |
Sadio Mané | 18 | Liverpool | Forward |
Matheus Pereira | 17 | West Brom | Forward |
Raheem Sterling | 17 | Manchester City | Forward |
Callum Wilson | 17 | Newcastle Utd | Forward |
Jack Grealish | 16 | Manchester City | Forward |
Roberto Firmino | 16 | Liverpool | Forward |
Jack Harrison | 16 | Leeds United | Midfielder |
Danny Ings | 16 | Burnley | Forward |
Dominic Calvert-Lewin | 16 | Everton | Forward |
Riyad Mahrez | 15 | Manchester City | Forward |
Chris Wood | 15 | Burnley | Forward |
Player | Squad | Age |
---|---|---|
Harry Kane | Tottenham | 27 |
Mohammed Salah | Liverpool | 28 |
Bruno Fernandes | Manchester United | 26 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2024 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Chang, V.; Sajeev, S.; Xu, Q.A.; Tan, M.; Wang, H. Football Analytics: Assessing the Correlation between Workload, Injury and Performance of Football Players in the English Premier League. Appl. Sci. 2024, 14, 7217. https://doi.org/10.3390/app14167217
Chang V, Sajeev S, Xu QA, Tan M, Wang H. Football Analytics: Assessing the Correlation between Workload, Injury and Performance of Football Players in the English Premier League. Applied Sciences. 2024; 14(16):7217. https://doi.org/10.3390/app14167217
Chicago/Turabian StyleChang, Victor, Sreeram Sajeev, Qianwen Ariel Xu, Mengmeng Tan, and Hai Wang. 2024. "Football Analytics: Assessing the Correlation between Workload, Injury and Performance of Football Players in the English Premier League" Applied Sciences 14, no. 16: 7217. https://doi.org/10.3390/app14167217