Steam Labs Search Experiment no longer takes tags too literally

For the longest time, Steam was really the only giant PC digital game distribution platform outside of publisher-specific launchers or the more niche (at least back then) GOG. With the Epic Games Store and even game streaming platforms nibbling at its market share, Valve has stepped up not only to redesign the way Steam looks but also the way it behaves. That has led to its Steam Labs Experiments that mostly focus on improving discoverability and, in this latest installment, how well it understands what you're searching for.

There are various ways you can come across new games you've never seen before on Steam. A lot of the focus has been on machine learning presenting you with titles it thinks you might be interested in based on games you've played or own but some people rely on searching based on certain tags. Unfortunately, Steam's current system for searching by tags is almost too literal.

Steam games have tags and some games have multiple tags. Unfortunately, some games have multiple tags that, to our human minds, should include related tags as well. We expect that looking for games with the RTS (Real-Time Strategy) tag will also bring up those with "Real-Time" + "Strategy" tags but, unfortunately, that's not how Steam's search works right now.

That's where Steam Labs Search Experiment 4.1 Query Expansion comes in. The mouthful experiment practically gives Steam's search engine a bit more common sense in relating certain tags to other tags or combinations of tags. All in all, it should at least solve some mysteries why games you see under "3D + Platformer" tags aren't under the "3D Platformer" tags.

Valve cautions, however, that this association isn't generic and won't make leaps based on one tag's common association with another. It's still far from being the smartest search engine for finding games. Maybe Google might still be a better fit for that.