How Algorithmic Bias Gives You Such Suck Teammates in Games?

Do you often find yourself paired with teammates who seem to lack skill in video games? Do you feel like your opponents are always disproportionately strong? Sometimes, this might not just be bad luck—algorithmic bias could be playing a significant role.

Algorithmic bias refers to unfair or discriminatory outcomes affecting certain target groups due to external factors like data or models during the design of algorithms. There are several common types of data bias:

1. Insufficient model or sample data: This results in skewed outputs due to inadequate representation in the data set.
2. Purpose-driven singularity: For instance, when Amazon analyzes your data to recommend products, it might not consider your actual purchasing power.
3. Systemic feedback loops: This often occurs in topics of high engagement, where systems repeatedly cater to your preferences, creating an “information bubble.”
4. Human bias: The design, adjustment, and application of algorithms can be influenced by the subjective biases of developers, users, or decision-makers.

In video games, matchmaking systems are a crucial element, often determining the competitive level and overall gaming environment. Several mainstream matchmaking models are currently in use:

1. ELO Rating System: Originating from chess, it adjusts player scores dynamically based on wins and losses to measure skill levels.
2. MMR (Matchmaking Rating) System: This system places greater emphasis on individual performance rather than just match outcomes.

It’s important to note that these algorithms were not biased at their inception. Game designers strive to balance factors like fairness, wait times, commercial interests, and competitiveness when developing matchmaking systems. However, when these priorities conflict, algorithmic bias can emerge.

For the long-term stability of their games, companies often avoid allowing all players to quickly achieve higher ranks. If players consistently lose, they may feel frustrated and quit the game. Conversely, if they win too often, they might lose interest due to a lack of challenge. To address this, algorithmic systems may deliberately match players with teammates of varying skill levels to control overall win rates. This regulated win rate can extend a player’s engagement by keeping them motivated to improve and try again.

From a commercial perspective, players who spend less in a game are more likely to be matched against those who spend more. In such cases, lower-spending players might find it harder to win, encouraging them to spend money on upgrades. Meanwhile, paying players are often given subtle advantages, such as being matched with weaker opponents, enhancing their satisfaction and increasing their spending.

However, matchmaking mechanisms have sparked significant controversy. For example, in Apex Legends, a poor anti-cheat system and fluctuating ranked matchmaking mechanics have led players to suspect that win rates are artificially controlled. This perception can erode trust in the game company, with players believing the game to be “unfair” or “manipulative.” Additionally, some players feel that games introduce excessive challenges, increasing frustration. A classic example is League of Legends, where many disillusioned players now prefer casual modes over ranked matches.

Reference

Laserface (2023). Matchmaking and Autofill. [online] League of Legends Support. Available at: https://support-leagueoflegends.riotgames.com/hc/en-us/articles/201752954-Matchmaking-and-Autofill.

Williams, K. (2021). Valve dev debunks Dota 2 players’ ‘forced 50% win rate’ theory. [online] win.gg. Available at: https://win.gg/news/valve-dev-debunks-dota-2-players-forced-50-win-rate-theory/ [Accessed 2 Dec. 2024].