The biggest project of Steam Labs, however, is the interactive Recommender. According to Valve, Steam's greatest asset is its extensive games catalog. But with so many titles, it's hard for customers to sort everything out and find the games that match their interests. That's why Valve introduces a new machine learning method that allows potential customers to benefit from the right games.
How the Interactive Recommender Could Solve the Steam Search Problem
Currently, Steam relies on a tag-based system that recommends customers titles-based games with similar tags added by users. However, the new Interactive Recommender will suggest games to customers based on the user's reading history and "other important data".
Valve indicates that the new system does not use tags or gender data at all, and explicitly collects only the game's release date. The Recommender learns the rest of the information it needs during the training process. According to Valve, not including the genre or the tags gives better results, but we will have to test ourselves the interactive recommendation of Steam.
Players will be able to modify their interactive recommendation by choosing to let the advisor focus more on newer or older games. and on the public or obscure titles. Valve says that its experimental recommendation tool will be able to help all customers, that they wish to have the latest versions of the game or hidden gems.
Valve also explained that, since the new Recommender relies on machine learning, the model can not recommend new games that customers have not played yet. Although the Recommender can learn quickly, Mr. Valve said that he kept the existing Discovery Queue system to help surface for new games. The Recommender will serve as a "complement to existing mechanisms rather than replacing them".
Independent developers say good recommendations on Steam are broken
The interactive recommender seems to be focused on solving Steam search problems. Especially with the way Steam seems to favor the big triple-A games at the expense of independent games.
In October 2018, Valve changed Steam's search algorithm, which added weight to sales and wish list activity when recommending games to customers. This resulted in a situation where the recommendations, in Valve's own words, "had the unintended side effect of defusing the tags in the" More of the Same Type "section on the game store's page." The result is that many independent game developers have seen drastic traffic reduction on the store page of their games.
Developer Jake Birkett was one of the first to report damage to their game sales through changes to the Steam algorithm. "In the past, I was optimistic about Steam, but these changes in discovery and recent changes in revenue distribution that only concern the most successful games do not make me particularly optimistic about the future of the game. selling games on Steam, "wrote Birkett on her personal blog. "In fact, I would go so far as to say that I am worried."
Although Valve says it has changed the referral system to address October bug issues, independent developers claim that video game sales have never fully recovered. Even as recently as previous sales of Steam Summer Sale, developers say that traffic from the Steam internal discovery queue has been lower since the changes made to the algorithm in October 2018.
Independent developer Yitz, Yitz, told Kotaku that confidence in Steam has been weakening since October 2018. "This Steam sale has been a disaster, but I am much more concerned about the general trend observed in the Steam." Steam algorithm since last October: push unpopular (including games "mainly negative" examined) triple-A games on titles for which Steam has enough data to know what would be better for the consumer. "
The new Interactive Steam Editor has just been launched, and is only available to users who have chosen to test the new experimental feature. But let's keep an eye on how this will potentially change the way games, especially indies, are discovered on Steam.
Matt Kim is a journalist at IGN. You can reach him on Twitter.