Decision Tree: Splits data over specific parameters (e.g., all teams shooting > 40% go one way). Fast to train.

Linear Regression: Maps each parameter to a coefficient linearly. Uses Ridge for classification.

K Nearest: Finds the K most similar historical matchups and uses their outcomes.

XG Boost: Fancy Decision Tree with gradient boosting. Great accuracy-to-speed ratio.

Random Forest: Collection of decision trees. Longer training but more accurate, provides probability values.

Score: Predicts each team's score using regression. Winner = higher predicted score.

Winner: Binary classification to directly predict the winner.

Note: Winner method performs ~6% better on average (73.5% vs 66.7%).

Select your Prediction Method, Model Type, and configure Hyperparameters.

• Prediction Method: Score (regression) or Winner (classification)
• Model Type: The ML algorithm to use
• Number of Models: Train multiple and keep the best

If unsure, Linear Regression + Winner is a solid choice!