The multiplayer online shoot ’em up Fortnite is the most popular video game in the world, with more than 250 million players.

It is also representative of a relatively new paradigm in gaming – one that has altered the software landscape until it has become almost unrecognisable in recent years. Known as games-as-a-service (GaaS), the game is funded through subscriptions and regular transactions by its players rather than a one-off fee.

Although this may not seem all that new to those of us old enough to remember the coin-guzzling likes of Pac Man and Space Invaders, GaaS means the continuous revenue stream funds ongoing development and improvement of the game.

And its cloud-based nature (along with the vast userbase) means huge amounts of data is available to developers Epic – the game generates a staggering two petabytes per month. All of this can be mined for insights into what keeps their players coming back for more – and contributed towards generating $3 billion in profit during 2018.

Data is the fuel of artificial intelligence. Every piece of data players generate by interacting with each other, and the platform can now be analysed with smart algorithms and used to make the game more entertaining and immersive, which in turn helps to grow the user base. 

These analytics – running on tools including Amazon Sagemaker – are used to identify KPIs such as the length of time players are engaging with the game – be that hours per day or sessions per week – and understand what it is that provides the “hook” that keeps players coming back.

Once this is done, the company can make changes to the game’s internal structure, such as deploying different gameplay modes encouraging players to interact in different ways, to drive growth across its key metrics.

Much of Epic’s infrastructure supporting Fortnite is built on Amazon’s AWS cloud service, and the company’s director of platform, Chris Dyl, told AWS Summit in New York “We use the data for everything, from ARPU (average revenue per user) to game analysis and improvements.”

At peak times, the Fortnite analytics pipeline ingests 40gb of data every minute into the Epic data lake, that as of last year was reported to stand at 14 petabytes in size, and to be growing by two petabytes a month.

The precise details of how data is used to improve gameplay and drive player engagement are kept under wraps. But tools and services related to data analytics that it is known to use include Amazon S3, the Apache Spark cluster computing framework, and the Tableau visualisation software.

As with other GaaS operations, it is likely that Epic use their data to balance gameplay within Fortnite’s “battle royale” tournaments – free-for-all skirmishes where 100 players compete to be the last man or woman standing in an ever-shrinking arena.

Additionally, design decisions such as the implementation of new gameplay modes will be based on what is learned about the types of battle that are popular with the userbase.

Because battles in Fortnite are relatively free-form, rather than being structured around missions or objectives as is the norm in video games, Epic can use analytics and AI to make interesting psychological observations around player behaviour. Will players tend to go it alone in a “lone wolf” fashion, or will they form temporary alliances and pacts to work together on self-determined objectives – such as taking down stronger players or holding strategically important terrain? What level of risk are players willing to take when it comes to exposing themselves or leaving cover, to collect better weapons and equipment? AI can be used to monitor this “emergent” gameplay and adapt the challenge to suit players’ style.

Emergent gameplay is nothing too new in gaming and marks a trend away from games where every individual element was hand-crafted by human game designers, towards allowing players to create rules for themselves in a “sandbox” environment. AI is a natural next step in this evolution, enabling the prospect of “virtual referees” that can adapt to what players are doing and ensure the game’s infrastructure adapts too, to cope with this behaviour.

Epic had also developed a great deal of its own infrastructure itself – way before Fortnite existed, its Unreal Engine was one of the most commonly used gameplay engines in both independent and “triple-A, ” big studio game design.

Carrying on this tradition, it recently made Epic Online Services freely available – a platform for creating, maintaining, and managing massively multiplayer, cloud-based games.

The move into the cloud and AI technology taken by Epic is driven by its partnership with Tencent, which invested $330 million in exchange for a 40% share of the company in 2012. Tencent is the most gaming-focused of the Chinese tech giants. It has pioneered the development of the cloud, mobile and microtransaction infrastructure rolled out by Fortnite, and now being adopted across the games industry.