04

NEWS

Online matching algorithms, big data, and gaming

Ever since its inception human versus human interactions have been at the center of video gaming. What began as simple ping-pong-type games over the last four decades has morphed into one of the most vibrant sectors of the entertainment industry. According to Statista, a digital economy analytical platform, the online gaming business will break the symbolic 100 bln. dollar cap by the end of 2023.

18 May 2022

Whether an online game achieves a sizeable fanbase and a commercial success often relies on players’ experience of how online matches (or battles, scenarios, campaigns, etc. depending on a game genre) between real persons are organized. When encountering a technical bug, glitch crash or even having to wait longer for an opponent, players tend to get impatient or irritated which can lead them to quit a game or uninstall an app. In such tense environments, every technical detail matters. It’s for that matter that running a successful online gaming company requires not only general business proficiency and accurate gaming industry insights, but also the ability to employ advanced mathematics and carry out data science research projects.

In order to better understand the role that matching algorithms play in online gaming, we sat down for a conversation with Runtian Ren Ph.D.,Warsaw University.

Let’s start with the basics. What are online gaming matching algorithms and why do they matter to, respectively, gaming platforms and gamers?

Runtian Ren: A matching algorithm can be interpreted as a mechanism, which lets players with similar skill levels play against or collaborate with each other in online games or applications. Millions of people choose mobile gaming (titles like, for example, PUBG) not only for entertainment, but also for social connections and person-to-person interactions. Nowadays, online player vs player games constitutes a huge segment of the entertainment industry. With so many players devoted wholeheartedly to online gaming on a daily basis, the gaming companies need the help of online algorithms to create gaming sessions for their users in an instantaneous manner.

The fact is, once a player sends a request to join a gaming session, he or she wants to battle with another player of roughly the same level as quickly as possible. If he or she isn’t assigned one gaming session in about 10 – 20 seconds, players tend to become impatient or angry, and as a result, stops using the app. Also, if he or she has been assigned to a session either against an expert or a rookie (compared with his or her gaming level), such a player may also become unsatisfied and decide not to play the game anymore. As such, for the purpose of maintaining a good reputation of the company, a good online matching algorithm is of great importance.

Can players hack or in other forms influence online matching algorithms?

No, not really. Players can’t have influence over the way gaming sessions are formed when they declare to join the sessions arbitrarily. Of course, there is an alternative: if one wants to play with friends, he or she can establish a session or server and have fun in a non-competitive manner.

So how does the problem look from the perspective of a company running a gaming platform or an app?

From the gaming platform’s side, the matching algorithms are the core of their business. Any company in the gaming industry that wants to maintain a good reputation needs to transform players’ gaming requests into high-quality gaming sessions. The latter is vital because it results in players being satisfied with their experience which consequently turns into a profit for the developer and the platform.

There are two central challenges that an online gaming matching algorithm has to address or balance. The first challenge is about the duration player has to wait to start from the moment when he or she sends a gaming request – the more time he or she waits, the more likely he or she becomes disappointed. The second challenge is about the skill level of the rival – the closer ranking between two players in a session, the more likely the players will enjoy the game.

As such, the platform has to, on the one hand, arrange good sessions with pretty much equal competitors, and on the other hand, it cannot allow players to waste too much time waiting for a “perfect match”. The problem starts when a player makes a play request and there is a shortage of similar requests since at that moment there are too few players of comparable levels of experience. In such an instance platform can delay the response to a player’s request assuming another player of nearly the same level shall appear shortly after and hence the problem might disappear naturally. But such a delay has a limit and players can get frustrated and quit the game altogether.

Alternatively, the platform can find another player of a different (lower or higher) level to form a gaming session. But that often leads to both players being unhappy about it, because you can’t really talk about competition when a beginner has to face a seasoned veteran gamer or even an e-gaming professional. In this fashion, the question of how to match the players into high-quality sessions becomes a real big deal both to the app company and its player base.

There are two central challenges that an online gaming matching algorithm has to address or balance. The first challenge is about the duration player has to wait to start from the moment when he or she sends a gaming request – the more time he or she waits, the more likely he or she becomes disappointed. The second challenge is about the skill level of the rival – the closer ranking between two players in a session, the more likely the players will enjoy the game.

Runtian Ren, PhD

postdoc researcher at Warsaw University

What happens when matching algorithms go wrong, how do players feel about such mistakes/failures?

Well, the matching algorithm almost always delivers – it can always produce a match. The issue is about the quality: if many players are disappointed about either the unbalanced rivals or time spent waiting for a response to their request response they are likely to drop the game or uninstall the app. This is of course unacceptable to the owners of the platform. They need as many players or users as possible, for it allows them to bring in more advertising revenue or directly charge players for playing.

What would you say are the biggest risks online multiplayer matching platform designers have to avoid?

The algorithm must be designed in an online situation. There is no credible information on the number of requests that shall appear in the future and the matchmaking decisions must be made at each second. As I said earlier, you also have to count in varying levels of players’ skills and the general tendency (particularly among teenagers who constitute a vast chunk of players) to get impatient instantaneously. So striking the right balance between speed (matching players of different levels) and quality (finding comparable opponents or teammates) is the key to business success in the online gaming world.

Which games/ gaming platforms have in your opinion the best matching models?

There are thousands of titles, but if I had to choose one particularly well-crafted I’d say Playerunknown’s Battlegrounds. It’s a very popular mobile game managed by Tencent Games, one of the biggest gaming companies in the world. It also uses the mechanism I have been describing before.

And what’s your experience with online gaming matching algorithms as a researcher? What interested you in the first place?

I had been previously working on some general and theoretical online matching algorithms with guaranteed worst-case performance, which is different from a practical situation that takes place in online gaming. In case of models I had been developing outside of the gaming industry, some easy-to-do algorithm usually does the trick. It classifies the requests into groups according to the player’s level and whenever two requests of the same level appear, it merges them into one session. As I already have said, such algorithms work in most cases, but tend to fail in a worst-case, real-life online situation.

In my approach I try to consider problems more practically (in a stochastic manner), i.e., utilize the distribution information of the arrival of the request at each time of the day to design the online matching algorithm. More recently, I’ve been thinking about proving whether these types of simple online algorithms work under such a stochastic model.

Runtian Ren received a BSc degree in mathematics and applied mathematics from the University of Science and Technology of China in 2014, and the Ph.D. degree in computer science from Nanyang Technological University in 2019. He is currently a postdoc at MIMUW. He mainly works on online algorithms.

Google Scholar: https://scholar.google.com/citations?user=CzlH3MsAAAAJ&hl=en

LinkedIn: https://www.linkedin.com/in/runtian-ren-720516125/

SHARE:

Skip to content