Table 1.
Algorithm performance across models (nested cross-validation).
Fig 1.
Algorithm performance comparison: classification vs. ranking.
The bar chart contrasts F1-scores and Capture Rates at Top 20% across five algorithms (LightGBM, Logistic Regression, SVM, Random Forest, XGBoost).
Table 2.
Gains chart performance by evaluation threshold (OOB bootstrap, n = 1000).
Fig 2.
Ranking efficiency and stability across thresholds.
(A) Cumulative gains curve with 95% OOB bootstrap CI and random selection baseline; (B) lift factor by evaluation threshold; (C) bootstrap distributions for Top 5% and Top 20% capture rates; (D) efficiency-coverage trade-off with the Top 20% threshold highlighted.
Table 3.
Feature importance consensus for goal scoring prediction.
Fig 3.
Feature importance consensus: process-oriented metrics vs. controls.
The bar chart displays the top 15 features ranked by multi-method consensus score on a log scale. Bars are color-coded by category, and annotations highlight dominant predictors such as Attempts at Goal, Total Offers, and Top Speed.
Table 4.
Tactical category importance analysis.
Table 5.
Position-specific top 5 performance factors for goal scoring (LightGBM).
Table 6.
Operational efficiency by evaluation threshold (OOB Bootstrap, n = 1000).
Table 7.
Scouting efficiency (OOB Bootstrap Mean, n = 1000).