Individual-Community Interactions and Emergent Strategies in League of Legends

League of Legends is a popular online video game where players battle against one another to climb in skill rank, like chess. Since a single player will encounter only five of approximately a million opponents in any single match, a player must use strategies that are better on average. They must have some understanding of which strategies will beat other strategies, similar to a social aggression hierarchy. Social aggression hierarchies have been well-studied in several species, such as primates, fish, parakeets, and insects [1]. In many of these species, the behavior of individuals is influenced by the overall aggression hierarchy structure as well as their individual awareness of it [1]. In League of Legends, players interact to form a directed network of strategies, which is functionally analogous to a social aggression hierarchy with strategies as nodes instead of individuals. This network is dynamic and strongly dependent on the game’s current rules, which are changed by the game’s developer every two weeks [2]. The interactions between players and game developers results in an open-ended evolution of strategies over time.
To better understand how certain strategies become popular via player-to-player interactions (assuming unchanging game rules), this project focuses on the relationship between the overall strategy network S and an individual player’s observation of the network s. A player’s observed strategy network s is determined by the sum of the observed strategy outcomes in their recent match history. Do players win more often if their network s is consistent with the overall network S? Players are split into two groups: Those who only use a single strategy and those who explore the strategy space. The former will be considered the control population since having a static strategy likely indicates they are trying to climb rank without navigating S, regardless of s. Players who have an accurate representation of S may be able to select better strategies, which may allow them to win more matches.
In animal aggression hierarchies, social strategies evolved over time periods longer than our observed data. With video game data, we are able to capture the creation, evolution, and extinction of several strategies under dynamic environmental constraints, which are very similar to non-digital social systems. This allows us to observe the entire player population history, strategy history, and the history of individuals. While the structure of S is solely determined by player-to-player interactions, the behavior of individual players is influenced by their knowledge of S. This feedback loop contributes to the emergence of successful strategies, broadly across different agents in social aggression hierarchies.