Application context based algorithm for player skill evaluation in MOBA games

This work is motivated by the challenge to improve player rating systems in Multiplayer Online Battle Arena (MOBA) games by incorporating the application context, as opposed to calculating the player rating based solely on the match outcome (winning or losing). The application context is modeled by using a set of parameters describing the players' in-game accomplishments and contributions to the team, as well as the performance of the team as a whole. The proposed application context based rating algorithm functions as an extension of the underlying (unmodified) player rating algorithm by adjusting the rating update step and can thus be applied for different rating systems in MOBA games. We evaluate the proposed algorithm by using a dataset comprising over 400,000 matches of the MOBA game Heroes of Newerth.