Valve changes Steam algorithms to give users more varied recommendations

Valve has released an update to its gaming platform Steam with the aim of providing users with a more varied and personalized offering in the game recommendations section.

Valve says it has not only made algorithmic changes but also fixed some bugs. This should lead to a range in the store that is ‘more precise and diverse’. According to the company, in the past, players found too few different games via the ‘More like this’ section and the recommendations, with the recommendations according to Valve also often showing the most popular games, without a degree of user-centric recommendations. .

The company behind Steam says it encountered bugs in the “Similar to tags” section of the recommendation feed, among other things; there, the top-rated games put too much of a stamp on what was shown to users. In addition, Valve reports that it found that in some places too short a timescale was used for calculating popularity, leading to an “unpredictable invisibility of certain games.”

To check whether the changes actually have an effect, Valve has already tested the changes in Steam with a small group of users. Recommendations algorithm adjustments and bug fixes have been tested on 5 percent of Steam users in recent weeks. According to the company, the users in this experimental group were found to be more likely to click on the games shown with the recommendations. This would happen up to 15 percent more often than in the control group. Valve says there was a higher degree of personalization and with it a greater variety in the range of games shown. The test group would also have visited 75 percent more unique games via the recommendation section.

Valve calls these results encouraging, stating that the “increased qualitative specificity and a wider range of titles on display really helped players find games they didn’t know they wanted.”