Dynamic difficulty adjustment (DDA) is a technology used to change a game's difficulty according to a player's skill. During a game, the difficulty adjustment technique may help the player to win if they are losing. In other cases, it can make it harder for a player to win a match.
What is Dynamic Difficulty Adjustment?
DDA monitors and predicts the length of time a player stays engaged by a game. It combines this information with different data types, such as how long a game keeps a player engaged in a single-player session.
A DDA can keep a player from getting bored if the game is easy. It can also keep players from getting frustrated if the game is too difficult.
DDA works in both the short-term and long-term. Short-term DDA prevents players from experiencing long stretches of the same outcome, whether good or bad. A random number generator is used to achieve short-term DDA. A long-term DDA adjusts the level of the game to one which is appropriate for their skills and performance.
That's all very well and good, but how does dynamic difficulty adjustment work in-game?
How Does Dynamic Difficulty Adjustment Work?
A common way of achieving DDA is to make changes to the course of the game by adjusting difficulty after trigger events occur that indicate undesirable states of the player. Such states include boredom and frustration.
DDA depends on machine learning algorithms to make the predictions needed to execute adjustments. Machine learning algorithms, such as supervised and non-supervised, create and update prediction models for games. Ensemble algorithms and instance-based algorithms are examples of logic used to create and update prediction models for DDA.
Systems for Dynamic Difficulty Adjustment
A patent granted to EA in 2018 reveals details of the technical components of DDA in EA games.
The patent describes a system with an electronic data store that a hardware processor uses to execute instructions to identify adjustment values to variables in the video game. The hardware processor generates a prediction model by executing instructions to access sets of data used in the machine learning system.
The patent also details how DDA uses different types of user interaction data to assesses how engaged a user is. Such data includes the amount of money spent in the game, the user's progress within the game, and the player's propensity to stop because of their in-game progress.
User interaction data is used in combination with other data types to create and act on gameplay prediction models. The data feeds different types of systems within the game that work together to change the difficulty.
The types of systems and processes that may work together include:
- Retention analysis
- Prediction model generation
- Cluster creation
- Cluster assignment
- Seed evaluation
- Difficulty setting
In a nutshell, these systems work together to collect player data, which is used to determine how difficult or easy the game should be.
DDA Data Modeling
The prediction model generation process involves historical user interaction data combined with control data to generate prediction models. The control data is used to set the desired prediction for the number of users.
A retention analysis system can be composed of one or more systems that generate retention rates and prediction churn for users. The predicted retention rate may be used to decide whether the game's difficulty needs to be changed. User interaction data is applied to prediction models to achieve this.
Users may be grouped into clusters based on interactivity data. Users who play the game for less than 30 minutes, for example, could be identified by the machine learning algorithm.
The patent suggests that in certain embodiments of the system, grouping users with similar characteristics and adjusting difficulty levels based on the unique actions of each user allows for better management of difficulty levels.
Cluster creation begins with the identification of users in the game. Data on user interaction is collected over time and used to filter out users who do not meet the interaction criteria. After the users are filtered out, user clusters are created with difficulty preferences based on the user interaction data and engagement levels.
Cluster assignment for a user is achieved by identifying the user and collecting the user's interaction data with the game over time. The user interaction data is used in combination with cluster definitions to identify specific clusters for users to associate with.
The difficulty setting process starts with the identification of a user, followed by the determination of a user cluster associated with the user. The configuration values are adjusted based on the user interaction data.
A seed evaluation system is used to determine how difficult a proportion of a video game can be. The seed evaluation process begins with the identification of seeds (values) that may be used to configure the video game. The progress of users for each seed is monitored over time to determine a difficulty based on normalized progress data.
A prime example of seeds is found in Minecraft, where different seeds provide completely different adventures.
In some embodiments of the system, the execution of DDA in the game may not be detected by the user. The game may also repeat changes in the video game if an event is triggered.
Why Does EA Own a Patent for Dynamic Difficulty Adjustment?
After discovering EA's DDA patent, many users of EA games became concerned about whether the technology was in use in their games and the effect it had on their experiences.
A lawsuit (which was later dropped) was brought up against EA in late 2020, giving rise to further discussions on the potential use of the technique by the gaming company.
The plaintiffs believed that EA used the technology to increase the difficulty of games so that more people would want to purchase in-game items (loot boxes) to win. EA provided information, and the prosecutors spoke to its engineering team to prove that there was no use of DDA or similar scripting as alleged.
According to an EA employee's announcement, the technology was designed to find out how to help players experiencing difficulty in games gain opportunities to progress. The intention is to ensure that payers do not get too bored or frustrated with the game.
EA delivered an official response:
We’ve heard your concerns around the Dynamic Difficulty Adjustment patent family (here and here), and wanted to confirm it’s not used in EA SPORTS FIFA. We would never use it to advantage or disadvantage any group of players against another in any of our games. The technology was designed to explore how we might help players that are having difficulty in a certain area of a game have an opportunity to advance.
EA stated that it would not use DDA technology to give or remove advantages for players in online games. It asserts that the technology is not in leading games such as FIFA, Madden, or NHL.
The Use of Dynamic Difficulty Adjustment in Video Games
EA has always denied using DDA in video games. Responding to a question on Reddit about DDA in FIFA, creative director Matt Prior stated that there is potential for player error in the game, based on individual player statistics and fatigue, rather than DDA.
It is not uncommon for patents in the gaming industry to be filed without ever being used. A significant amount of research and development goes into creating new concepts for gameplay. New ideas are always generated that may not get off the ground due to different factors, such as reputational risks or even just not finding a way to properly integrate the idea into a game.