Networks in skill-based matchmaking in online video games
https://www.semanticscholar.org/paper/Matchmaking-for-online-games-and-other-P2P-systems-Agarwal-Lorch/91907feaf8dfb0bda37781cf1c17a961d50960b7
https://comicbook.com/gaming/2018/04/25/league-of-legends-positional-matchmaking-separate-ranks/
Competitive multiplayer games face the challenge of providing their entire player base with a fun and balanced play. However, the population of players can vary widely in skill level, and when a new player competes against a seasoned veteran, neither player has a fun experience; the experienced player does not engage himself in a competitive challenge and is soon bored, and the new player is utterly defeated, easily losing their will to play more games. The game is at its most fun for the participants if the outcome is uncertain in the sense that each of the participating teams has a fair chance of winning.
The producers have a financial incentive to keep their players interested and excited to continue playing the game, and therefore would want to match players based on their skill level within a player versus player experience. Most online games use the classical Elo system, which calculates the relative skills of players and adjusts players’ ratings after the results of each match. The popular multiplayer online battle arena video game League of Legends (LoL) uses an implicit Elo system in which the actual Elo score of players are hidden; instead, players are assigned to ranks and tiers, such as Diamond 2.
Once players queue to enter a game in League of Legends, the matchmaking system selects 10 players in 2 teams of 5 with similar Elo ratings to enter a game together. If a networks model was applied to this matchmaking system, it would resemble a large network in which players are nodes. Based on a predetermined limit as to the maximum allowed difference in Elo rating that is allowed for two players to play a game together, an edge between two players would represent the possibility of playing a game between two players whose difference in Elo rating is below the threshold. If two nodes are not connected, it would mean that their Elo ratings are too far apart for a game to be played between them given the threshold. The system would then identify a component of the network of 10 nodes; since there are edges between every node in the component, every player is within the acceptable range to play a game with each other. To keep the queuing time reasonable, the threshold in ratings may be dynamically altered based on how many players are online and in queue.
A consequence of this system is that queue times are longer in different conditions. As with most games, League of Legends has a large player base whose skill level distribution resembles a bell curve. Therefore, the best of the best players experience extremely long queue times since there are less players who are at their level. The waiting time increases further since each match takes around 40 minutes and so only a small proportion of players are in queue waiting to enter a game. A solution to this would be to increase the skill threshold to include lesser skilled players, but this circles back to the original problem in which high skilled players are matched with players who are vastly inferior.
Skill based ranking and matchmaking are crucial elements for designing a competitive and enjoyable online gaming experience. Therefore, game designers must meet a careful balance between competitiveness and convenience for its players.