Last July Steam launched a new interactive recommendation system to suggest games to its users. First limited to a simple Steam Labs tool, it has just been extended as a real functionality integrated in the shop.

Concretely, AI using machine learning, will be able to offer you games that you should enjoy based on your shopping habits and consultations, but also the practices of people with similar tastes.

A very reliable recommendation system

Steam has detailed on its site the functioning of its new tool which seems very promising:

The interactive recommendation system uses a machine learning model that has been trained from the playing time history of millions of Steam users. The algorithm is not directly influenced by tags or ratings. Rather, it is based on the games that users actually play to build their knowledge of games on Steam. The model is based on the idea that if players with game habits that are broadly similar to yours also play a game that you have not yet tried, then you are likely to enjoy this game too.

You can therefore filter your selection using tags, decide to search only among the most popular or niche games, old or classic, in order to find the rare pearl. In any case, a quick glance allows you to judge the effectiveness of this new tool. Suggested games generally meet expectations. For example, the recommendations that were made to me were not yet on my wish list, but I had plans to buy them soon.

This new recommendation system is accessible from your Steam software but you can also go there directly on the Internet here.