Soccer Analysis Dashboard
Penalty Direction Prediction Demo
Interactive demo of penalty direction prediction
Penalty Direction Statistics
Distribution of penalty kick directions (Total: 3,540 penalties)
| Direction | Count | Percentage |
|---|---|---|
| Left | 2113 | 59.7% |
| Right | 1427 | 40.3% |
Penalty Direction Prediction Methodology
Overview of the approach used for predicting penalty directions
- Problem Definition: Predict penalty shot direction (left or right) based on various features.
- Dataset Analysis: 3,540 entries, slightly unbalanced (59.7% left, 40.3% right).
- Metrics: F1 Score and Matthews Correlation Coefficient (MCC).
- Data Splitting: 64% training, 16% validation, 20% testing.
- Model Selection: Tested Logistic Regression, Random Forest, Gradient Boosting, and SVM.
Model Performance on Validation Set
Comparison of different models' performance
| Model | F1 Score | MCC |
|---|---|---|
| Logistic Regression | 0.10 | 0.014 |
| Random Forest | 0.36 | -0.003 |
| Gradient Boosting | 0.35 | 0.045 |
| Support Vector Classifier | 0.00 | 0.000 |
Best Model Performance (Random Forest)
Test set performance of the selected model
The Random Forest model was selected based on its performance on the validation set.
F1 Score
0.43
MCC
0.077
These results suggest that further improvements could be made through hyperparameter tuning, feature engineering, or more advanced models.