BUS612: Data-Driven Communications

League of Legends is a multiplayer online battle game in which two teams fight each other to destroy the respective enemy base. We collected data related to League of Legends matches and player performances with the aim of extracting meaningful information about human behavioral patterns. For this purpose, we took advantage of Non-negative Tensor Factorization (NTF), a technique that allows to extract correlation in the data in several dimensions at a time. The advantage of using such a technique lies indeed in the opportunity of disentangling the topological and temporal characteristics in the data, and exploring and validating them separately.

Here, we analyzed a dataset composed by nearly 1000 players, characterized by different features, e.g., number of kills, number of deaths, etc., which varies over time, from match to match. We represented the data as a tensor and we applied NTF to extract the factor matrices related to the players, the features, and the temporal activities.

The analysis of the NTF outcome and the application of clustering methods, such as the k-means, highlighted the presence in the data of several groups of players, characterized by a correlated behavior in time and topology. In particular, players belonging to the same component (cluster) are characterized by similar features and activation over time. We carried out the analysis of the topological characteristics of each group of player by looking at the features highlighted by NTF, and comparing the interpretation derived by these results with the original data. We found good agreement between the NTF output and the characteristics of the discovered groups of players in the original data. Therefore, NTF successfully identified groups of distinct behaviors in the data that can be interpreted as different player strategies.

It would be expected that different strategies (e.g., collaborative vs. individualist playing) would lead to diverse performances (e.g., affecting the winning/losing ratio). However, by investigating the temporal activity patterns of the player groups, we found that they are mainly characterized by a constant behavior that is active continuously over time. Thus, the analysis of the temporal activation of the NTF components stressed the reluctance of players to adapt their strategy and gaming behavior over time.

This finding might be due to the game design of League of Legends: the team formation in the game is based on a mathematical rule which aims at contrasting teams with comparable skills, thus yielding the same prior probability of victory to each team.

We supported this fact by computing the Kernel Density Estimation over the feature winner for each player, divided by clusters. Only marginal, yet statistically significant, differences emerged, which are likely not perceivable by the players. Thus, players are not incentivized to change their strategy with another one.

In conclusion, the techniques and approaches used in this work are promising, and open new questions about human behaviors in multiplayer online games. The information provided by our framework could help game developers to design recommending systems for users and enhance users' engagement. Uncovering players' strategies would allow to recommend different types of champions, such as "support" champions if players are characterized by assisting actions, or "action" champions if they are otherwise more inclined in killing. Moreover, monitoring the group each user belongs to over time could help in promoting engagement in the game, through the use of custom rewards.

Future work will be devoted to the analysis of both different aspects of the game and of additional game datasets with the aim of exploring behavioral patterns in different scenarios. We are indeed interested in verifying if player strategies change depending on the champion player select and how different roles affect their performance over time. Moreover, we would like to study if the characteristics identified by the NTF might reflect the strategies of players with different skill levels. We are working on reproducing the evolution of skill level for players in LoL by computing a proxy, namely the TrueSkill (https://www.microsoft.com/en-us/research/project/trueskill-ranking-system/), for the actual score used in the game, which is not publicly available in the game data. We also aim to investigate, given the possibility of disentangling player's behaviors over time, if it is possible to nudge players to change their strategies by the use of incentives, such as high percentage of victory. We finally plan to test other tensor factorization techniques, such as PARAFAC2, which will allow to integrate players having variable numbers of matches in their gaming history.