How personalization unearths the perfect gaming match

Electronic Arts' director of data science Scott Allen explains the importance of personalization for games, in particular its use in creating archetypes for the perfect matchmaking experience


Staged on the backdrop of a hyper-competitive, demand-driven industry, an awful lot goes into building an engaging game today. It is, therefore, no surprise that data is crucial to ensuring a games' success. But, for Electronic Arts (EA) director of data science Scott Allen, the most important facet of data in gaming is its ability to help us personalize games to create the most satisfying and challenging experience for gamers.

Personalization is beneficial for every industry, with 74% of consumers having chosen, recommended or paid more for a brand that provides a personalized service or experience, while 74% report feeling frustrated when website content is not personalized, according to Infosys.

But for a sector with as many individual, data-generating events as gaming, it is essential that companies demonstrate an understanding of their users to create lifelong customers. And, as Allen puts it during his presentation at DATAx San Francisco, "personalization is first about understanding uniqueness".

According to Allen, EA's approach for catering to its 500 million players is to use their unique data to categorize their behavior into five archetypes, each with distinct approaches to gaming. He shares this example from Battlefield 1.

"If we understand how people like to play, that helps us out with a lot of their problems," he explains. "For example, if you have a 15 vs 15 or a 10 vs 10 game that already contains a heavy elite seeker and a suppression master, we would not put new elite seekers into that game – they just won't have fun."

Through using archetypes, EA can personalize games so that they are balanced and contain the right match for all the players, thereby enhancing their gameplay experience.

Allen then touches on another very common use of personalization in gaming, recommendation engines which present users with options they think they will enjoy based off the data the company holds.

"There are two different kinds of recommendations: One where content is just content and the other where the content is a person," he explains. "If you look at it that way, any sort of matchmaking, any sort of friend that algorithm finds and any sort of toxicity measure is also a recommendation. That recommendation system is created by utilizing the affinity of confirmed actions."

This means that gamers' data can be used to pair them with other gamers who are going to form the best fit – and the best game. He gives the example of EA intentionally not placing a passive player who needs support with another passive player because "it's just not a good experience for them".

And this method has been a breakout success for the company – and its players. When, a few months into its introduction, EA employed matchmaking in its breakout game Apex Legends it improved retention massively. In fact, the company has found that a more balanced game can raise retention by upward of 20%.

This ability to improve positive experiences in games has also helped EA to reduce toxic player behavior that ruins games for the other users.

"Toxicity in games doesn't just come down to what they are saying, it is also about what they are doing," explains Allen. "They can throw off a game and create an enormously negative feeling for everyone."

One particular method for dealing with toxic players is his favorite: Matchmaking them in games together so they cannot ruin it for anyone else (an approach the room murmurs in approval of).

"There is a significant opportunity right now for personalization in games," he concludes. "We are able to offer dynamic experiences and we are able to change the game while people are playing it. To do that we have to be able to get these insights and get these behaviors modeled out correctly in a way that is both explainable and actionable. And we have to be able to cohesively pull together and understand methodologies that can be worked around and then process a system that works best for everyone."

Five criteria for evaluating managed sd wan solutions small

Read next:

Five criteria for evaluating managed SD-WAN solutions