Valve is launching a new “Interactive Recommender” that generates personalized lists of Steam games users might like based on what they’ve played before. It’s not replacing the platform’s existing recommendation systems, but it’s supposed to give users another way to navigate Steam’s massive backlog of games — especially smaller titles that often get lost in the mix.
The Interactive Recommender uses machine learning, which Valve notes “all the cool kids are doing,” for what sounds like a collaborative filtering system. Instead of focusing on tags or metadata, it “looks at what games you play and what games other people play, then makes informed suggestions based on the decisions of other people playing games on Steam. The idea is that if players with broadly similar play habits to you also tend to play another game you haven’t tried yet, then that game is likely to be a good recommendation for you.”
Once that’s done, users can narrow down the recommendations with tags and sliders. So you can browse only for categories like “Massively Multiplayer” or “Indie.” You can include only newer releases (the slider moves from 10 years to six months) or you can weigh the results from “popular” to “niche.” The last feature works similarly to Last.fm’s music recommendation system — to which, overall, the Interactive Recommender seems similar.
Valve describes this section as “a very effective way to find hidden gems.” My niche results weren’t incredibly obscure; Steam recommended Void Bastards and Layers of Fear 2, for example, among many other titles that have gotten mainstream games press coverage. But at the very least, most of them were games I hadn’t paid much attention to and probably wouldn’t have thought to hunt down. And they were actually things that looked interesting to me — unlike most of the games in my normal Steam discovery queue.
The biggest drawback is that at least so far, you seemingly can’t remove items from the recommendation list, even if you’ve already added them to your wishlist or opted to “ignore” them on the store page. So unless you buy a lot of the games — and specifically buy them on Steam, so they’ll be in your library — you might get diminishing returns over time. This limitation also makes the “Popular” results much less useful, since mine included several games I already own on other stores and platforms.
Valve notes that the Interactive Recommender doesn’t help you find new games either, since nobody is playing them yet — although it supposedly picks up releases “with just a few days of data.” That’s part of why Valve isn’t changing the existing recommendation options, which are based on different models and data. You’ll still see the same new, trending, and popular games on the main page, and you can still browse recommended tags, plus all the other curated selections and lists that Valve offers.
The Recommender is part of a new initiative called Steam Labs, where Valve will let users check out experimental features. (Don’t confuse it with The Lab, a Valve-made series of virtual reality experiments, or Aperture Hand Lab, a demo for the Valve Index VR headset.) Steam Labs features two other projects right now: Micro Trailers, which is a large directory of six-second game trailers, and The Automated Show, which is a half-hour video compilation of new Steam game trailers. Valve also references an intriguing project codenamed “Organize Your Steam Library Using Morse Code,” although it sadly doesn’t make an appearance.
While Valve is pitching these features to its users, the Interactive Recommender seems like a nod to smaller developers who are frustrated by its storefront. Valve’s recent Steam summer sale was reportedly a letdown for many indie creators — partly because of some confusing contest rules that led users to cull games from their Steam wishlists, and partly because developers felt Steam was promoting large mainstream games at the expense of more niche titles. Valve also promises the Recommender won’t force developers to game the system with specific tags, pricing, or advertising styles, although without knowing more about how it works, it’s hard to rule out users discovering quirks in the engine.
Valve tends to push mechanical, data-based solutions to Steam store problems, sometimes in situations where it seems inappropriate. (See, for example, its review-bombing charts.) But here, “more data” is exactly what people need to get beyond Steam’s front page — and remember that there’s still a ridiculous amount of cool stuff on the platform.