A comparison of published models performing outcome prediction of League of Legends matches. We have excluded studies where no test split is specified and whose dataset size is smaller than 1000 matches. Marker sizes are scaled by the logarithm of the dataset size.

Model timing indicates the point in the game up to which the model was trained to make predictions. A timing of 0 indicates a pre-game prediction. A timing of ∞ indicates a post-game prediction.