Soonchunhyang University

Action2Score: An Embedding Approach to Score Player Action

Action2Score introduces a sophisticated deep learning embedding model designed to quantify individual player contributions within Multiplayer Online Battle Arena (MOBA) games.

• The Action2Score model achieves over 99% discernment accuracy in evaluating individual player contributions by analyzing action sequences rather than binary match outcomes.
• The system utilizes a GRU-SLP architecture trained on a dataset of 245,575 League of Legends matches to map player behavior into a 30-dimensional embedding space.
• By decoupling performance from final match results, the model effectively identifies the value of support roles that are typically undervalued by traditional metrics like KDA and gold.