One more level: Data science in the gaming industry

Are you a gamer? I’m not but in my teens I could spend hours and days glued to strategy games. Someone is doing a great job creating engrossing masterpieces that make us play “yet another round”. Did you know that nowadays data science plays an important role in game production? Right, it’s not only about breathtaking design, twisted plot and endless opportunities to level up your character.

I’ve talked to Mazen Aly who is a data scientist at King — a company creating online games. Mazen is an EIT Digital alumnus with a master degree in Data science from KTH in Stockholm and TU/e in Eindhoven. Let’s take a sneak peek behind the curtains and learn how data analysis helps design even more exciting virtual adventures.

Mazen, what kind of purposes do you as a data scientist strive to achieve working for a mobile games company?

For a data scientist in the gaming industry, the ultimate goal is create better games that are fun to play. We have make sure that players enjoy their time while playing. This is done thanks to collaboration with the game teams aimed at creating better features. Another goal is to improve the key performance indicators like acquisition, engagement and monetisation metrics.

What kind of data do you collect and, respectively, which aspects of user behaviour do you analyse?

We collect historic data of the game rounds to help us understand the players’ behaviour and determine which aspects of the game the players generally like or dislike and thus, improve and build better games based on these insights. One simple aspect to look at is the number of game rounds played in a certain level. If players are playing a certain level more often, the root cause is analysed to understand why this change is seen.

How do you tie in the results of your analysis with product development? What impact do your insights make on the product?

Many decisions in product development are based on data analysis. Let us say we are working on a new feature and there are two ways to develop it. We can’t just pick one of them by gut feeling. One way to make a decision is to run an A/B test which is a powerful tool for product development as it helps estimate the value of a new change to the product while reducing the effect of some confounding factors like seasonality effects.

You mentioned user engagement. Can you give me a specific example of machine learning to help increase retention rate and reduce user churn?

Machine learning can help in estimating the players’ likelihood to churn. This is important in assessing the quality of new games. If most of the new players in a new game have low probability to churn that means that the game is good in this aspect. Moreover, estimating a player’s probability to churn can help us take actions to avoid that. This is done by using training data or historic data of the players’ behaviour in the game and applying a machine learning classification algorithm to estimate the probabilities.