Current Model Performance
Diagnostics for the 5 models behind our predictions: current-season accuracy, trained hyperparameters, confusion matrices, and confidence distributions. For weekly accuracy tracking, see Past Performance.
Loading model metrics...
Goals Market Visualizations
Model Comparison (Test Accuracy)
Probability Distribution (Over 3.5, All Models)
KNN Confusion Matrix (Over/Under 3.5)
XGBoost Confusion Matrix (Over/Under 3.5)
TabPFN Confusion Matrix (Over/Under 3.5)
LightGBM Confusion Matrix (Over/Under 3.5)
CatBoost Confusion Matrix (Over/Under 3.5)
Match Winner Visualizations
Model Accuracy Comparison
Confidence Distribution
KNN Confusion Matrix (Match Winner)
XGBoost Confusion Matrix (Match Winner)
TabPFN Confusion Matrix (Match Winner)
LightGBM Confusion Matrix (Match Winner)
CatBoost Confusion Matrix (Match Winner)
About These Models
The 5-Model Lineup
KNN, XGBoost, TabPFN, LightGBM, and CatBoost are each trained independently on historical Premier League data to predict Over/Under goals and match winners. Their predictions are combined into an accuracy-weighted ensemble for the current week's picks.
View Interactive Predictions →Reading the Diagnostics
The confusion matrices show how accurately each model classifies matches, and the probability/confidence distributions indicate how confident each model is in its predictions. These are computed on each model's held-out test set from the current training run.
Weekly Accuracy Tracking
This page reflects the latest training run's metrics and is updated whenever the models are retrained. For how the models have actually performed week over week against real results, see Past Performance.