Results
The analysis and predictions generated by the models have been evaluated using metrics such as Mean Squared Error (MSE) and R-squared (R²). Below are the key findings and results:
Top Goal Scorers Prediction
R²: 0.5394
MSE: 4.7182
The model accurately predicted the top 10 goal scorers for the 2023-2024 season. Players like Lionel Messi and Karim Benzema were among the top performers.
Top Assist Leaders Prediction
R²: 0.9715
MSE: 0.2752
Predictions for assist leaders highlighted Antoine Griezmann and Alex Baena as top contributors, showcasing the model's high accuracy.
Clustering Analysis
Discipline, workload, creativity, and goal-scoring clusters provided insights into player groupings:
- Discipline-Based Clustering: Silhouette Score: 0.62
- Workload-Based Clustering: Silhouette Score: 0.54
- Creativity Clustering: Silhouette Score: 0.51
- Goals vs Assists: Silhouette Score: 0.64
Match Win Probability Prediction
The win probability model predicted match outcomes with:
R²: 0.8707
MSE: 1.1809
Visualization of predicted probabilities:
Best 11 Players Model
For each season, the model identified the best 11 players using performance metrics such as goals, assists, and minutes played.
- 2020-2021: Lionel Messi, Karim Benzema, and Gerard Moreno ranked top.
- 2021-2022: Karim Benzema, Vinícius Júnior, and Iago Aspas were standout performers.
- 2022-2023: Antoine Griezmann, Robert Lewandowski, and Karim Benzema led the list.
- 2023-2024: Artem Dovbyk, Alexander Sørloth, and Robert Lewandowski topped the rankings.
Player Classification
A Gradient Boosting Classifier categorized players into roles such as:
- Elite Goalscorers: Players with high goal-scoring efficiency like Karim Benzema.
- Playmakers: Top creators like Antoine Griezmann.
- Aggressive Defenders: Defensive stalwarts with impactful metrics.
R²: 0.9694
Conclusion
The project successfully demonstrated the power of machine learning in analyzing football player and team data. The models provided actionable insights for:
- Player Performance: Identifying top scorers and assist leaders for strategic planning.
- Team Recommendations: Suggesting players to enhance team dynamics.
- Clustering: Grouping players to reveal patterns in discipline, creativity, and workload.
- Match Prediction: Accurately estimating win probabilities for informed decision-making.
- Player Classification: Categorizing players based on performance to inform strategy.
These findings emphasize the potential of data-driven insights in improving team performance and player management strategies.