Steam and Discoverability

This is not about advice for getting your game discovered! This is a philosophical argument against the way the platform works right now.

I’ll preface this by saying that I am not the most knowledgeable person on this topic, and I’m sure I’ll get some things wrong. This is all based on my understanding and what I’ve heard from Valve reps at things like Dev Days and feedback sessions. Steam’s algorithms are also secret, of course, and change from time to time, so some of this is implication or guesswork.

If you’re interested in the state of discoverability in game marketplaces in general, with actual usable advice, I can’t recommend Simon Carless’s newsletter enough: https://newsletter.gamediscover.co/

Basically, Valve has a number of ways that they attempt to surface new games to players. Among others:

1. If a player is on Steam, they’ll see popups from their friends, telling them what their friends are playing right now.
2. When a developer updates a game, Steam has a number of tools that help push that game to new audiences based on updates, including Announcements and Update Visibility Rounds. Valve recommends updating frequently, pushing major updates every 1–2 months and minor updates more frequently.
3. Steam has a “Discovery Queue” that pushes “the most popular recently-released titles” in the store. (https://www.gamasutra.com/blogs/ErikJohnson/20190404/340061/A_Deep_Dive_Into_Steams_Discovery_Queue.php)
4. There’s also a “Recommendation Feed” that recommends games based on other titles the user has been playing, as well as a host of other algorithmic factors

On the face of it, none of these things seem unreasonable. It all sounds like good ways to decide whether someone would like a game, if you are using algorithms and data-collection to do so.

But, if we dig a little deeper, we’ll notice something about the types of games that thrive most under this system. First, anything that people play a lot, or for a long time, will both get algorithmic boosts in “popularity” as well as being shown to your friends more often. Second, any type of game that updates frequently will have an advantage, both algorithmically and in terms of access to the Update Visibility. Third, already-successful titles get boosted more. Fourth, titles that have a lot of positive user reviews get a boost. And fifth, titles that are similar to other things you’ve been playing get help.

All of this points to types of games that Valve encourages sales of, and by extension, encourages the creation of. Titles with a lot of gameplay hours, either because they’re long games or things like multiplayer titles that get returned to often. Titles that have frequent updates — anything that falls into the “Games as a Service”(GaaS) idea (which is a philosophy promoted by Valve, both in their developer-facing materials and in their own development of games: https://www.youtube.com/watch?v=Fwv1G3WFSfI). Successful titles will get an additional boost by virtue of their success, so any games that appeal to a wide audience. This also feeds into the review metric — titles that review well among a broad spectrum of people will succeed. (Because this seems obvious, I’ll point out that reviews are not solely a measure of quality, but also about existing user expectations and how well the game matches that). And lastly, games that are similar to other successful games will get seen more often.

We can flip that construction on its head to figure out what sort of games get the least help from Steam. That’s not to say that these games can’t sell on the platform, but they are the sort of titles that need to work harder to do well. The worst type of game would seem to be single-player, short, not updated with new content frequently, appealing to a niche audience, or experimental and unlike any type of game that already exists.

Now, I think it’s very possible to look at these criteria and think they make a lot of business sense. Especially if you’re thinking, “how can Steam push things to users using an algorithm in the best way possible?” Algorithms have to use quantifiable data to make decisions; things like playtime, revenue, numerical review scores, update frequency, and player history are all things that can be measured by a computer and used in an algorithm.

But: is this the world that gamers, developers, and platforms would LIKE to exist? Again, by playing kingmaker here Valve not only influences what does well in the current marketplace, but also what gets made in the future. They are not just recognizing the rise of GaaS, but also promoting it. Personally, the kind of games I love to play and make are generally small, unique, narrative-focused, and do not have broad audience appeal — exactly the sort of title that gets no help from Steam’s algorithms. Even if you don’t personally enjoy those types of games, I think it’s clear that new types of genre, audience, and experience are only discovered through experimentation, usually by small teams aiming at a niche audience (because they have to find their own audience for a new thing!). And so I think that the vibrancy and health of the medium depends on those types of games being at least somewhat successful.

So I would argue three things: First, an algorithmic approach to discoverability will always have blindspots and be exclusionary. Relying on quantifiable numerical data will miss huge swathes of what people enjoy about games (which are art and thus entirely subjective). I know that for me, the games I have enjoyed the most and have had the largest impact on me are not the ones that show up in my Most Played list.

Secondly, the criteria which Valve has chosen for Steam’s discoverability rewards certain types of titles and punishes others. It’s obviously possible for weird, story-driven single-player games to be successful in this universe (Disco Elysium springs to mind), but the deck is at least somewhat stacked against them. (And this takes place in an environment where small, story-driven, experimental games pay the same 30% cut to Valve as do the games that benefit strongly from the discovery ecosystem.)

And third, that this systemic bias towards certain types of games and away from others harms the medium as a whole by penalizing experimentation, diversity of experience, and smaller indies (who do not usually have the ability to enter the GaaS realm).

Steam is a business, and a very successful one. I can’t present a compelling short-term business case why they should change discoverability to help small, weird, niche titles. The best I can do is gesture vaguely at the long-term health of games if we lose new types of genre, small teams, and a diversity of business models for games.

But I hope this helps make a case for why the current status quo may not be ideal for us as game creators, players, and people who love games in general.