DraftComPromise - On Draft Composition Recommendations in League of Legends

Multi-player online battle arena games constitute a highly popular category of online multi player strategy games. In League of Legends, a game with one of the most active player communities globally, ten players are divided into two teams, each selecting a unique champion-character after another during the so-called draft phase. With over 150 champions in the game to choose from and each possessing distinct synergies and abilities, choosing suitable champions for the constantly changing draft situations can be challenging for players. In this paper, we explore how draft recommendation tools work and investigate a draft's key factors by means of statistical analysis of a potential champion's impact on the match. Based on the an-alyzed factors, we propose concepts for a draft recommendation. We implemented our concepts as a real-time recommendation software - DraftComPromise. We utilized DraftComPromise to evaluate our concepts by means of a comparative study using recommendations generated by our tool and comparing them to those generated by four existing draft tools. The results of our study revealed that the established draft tools for LoL and also DraftComPromise were close together in terms of both statistical analysis regarding the win rate but also the choice of champions that were recommended. DraftComPromise performed better than three and worse than one of the established tools within our study, however, the margin of error values indicate that none of the tools was able to recommend champions for a composition that will statistically win over 50 % of the time.