You’re going to see a more personalised list of Steam recommendations now, as Valve has rolled out some changes to the algorithm which will more heavily prefer games you might actually like, instead of just the titles that are more popular. You should be seeing these changes take effect everywhere Steam recommends games as of today – just ahead of the new Steam library beta.
Valve’s vague about exactly how these changes work – likely in an effort to prevent exploits of the system – but one of the changes is an expansion of the time period used to calculate popularity, which previously created “unpredictable visibility for some games.” There are some bug fixes in all this, including a bug where the ‘similar by tags’ was getting more heavily influenced by top-rated games than intended.
The new system already rolled out to around 5% of Steam users over “the past few weeks.” Users in the experimental group were 15% more likely to click on recommended games.That drove an overall 75% increase in unique games visited, and a 47% increase in unique visits per game – all of which suggests this is broadly positive for both players and developers.
“The increased personalisation means there is an even greater variety of games being shown in this section,” Valve says, “and customer impressions are more evenly distributed among them.” Thus, the experimental rollout was judged successful enough to reach Steam as a whole.
Valve says it will “continue to make changes and run experiments like this in order to improve Steam’s existing features,” along with more formal efforts like Steam Labs and the Interactive Recommender. Hopefully, this also means we can expect more detailed explanations of impending changes, too, since a lack of communication from Valve has been a common complaint, especially among developers.