Jump to content
Sign in to follow this  

Valve Rolling Out Interactive Recommender for Steam

Recommended Posts

With the sizeable catalog of games Steam has to offer customers, it can be a daunting task to decide which ones may be worthy of your money and time. Reviews and tags can help you determine if a game is to your tastes, and now Valve has developed a new, experimental tool that may also prove useful. The Interactive Recommender leverages a neural network to recommend games to you, and you can control parameters like the game's popularity and age to suit your desire. You can also filter the list to only show games with a selected tag, or to remove games with a selected tag, but tags do not feed into the neural network.

As Valve explains in the post introducing the new tool, the network does not look at anything extrinsic to the game, like its reviews or tags on Steam, or any other kind of special analysis of the game. Instead it just looks at the play patterns of users, and its suggestions to you are based on what people with similar play patterns are playing. This does have the issue that it can take a few days before new releases can be picked up and recommended by the tool.

Because the Interactive Recommender does not use algorithms that rely on extrinsic data, it should be at least difficult for developers or publishers to manipulate it. The amount of time spent playing a game is what feeds into it, as well as the game's release data, so for a developer to want a game to appear higher, it needs to be one people want to play a lot of.

Currently the Interactive Recommender exists as an experiment and is meant to be used in combination with other tools, like the Discovery Queue, for finding new games to try. It may prove somewhat interesting to see how it develops. (It might also be interesting to see if turned inward, and recommend what game already in my library I should play next.)

Source: Valve

Back to original news post

Share this post

Link to post
Share on other sites
Sign in to follow this