How Steam's Discovery Algorithm Actually Works
Updated August 5, 2026
Steam’s discovery algorithm is a reactive ranking system: it surfaces games according to what players are buying and playing this week, and it makes no advance judgement about which releases deserve an audience. Valve put that plainly in a 2023 developer talk at devcom: “Instead of predicting success, we’ve learned to follow player interest,” said Erik Peterson of Valve’s Steam business team, “the games they are excited to get into and play, right now.” That single design choice explains almost everything else, including why so many good games never reach you.
It also means most of the advice you have heard about gaming the algorithm is aimed at metrics Valve says it ignores.
What does Steam’s algorithm actually measure?
It measures whether people buy a game and keep playing it. Valve’s Steamworks visibility documentation puts the question to developers directly: “Are people buying and playing your game? This is a strong signal to Steam that there is interest in your game and that Steam should try showing it to more players.” Revenue is the primary reading of that interest.
Everything the store shows you flows from that. The main capsule on the front page of the store, in Valve’s words, “is driven by what players spend time and money on, rather than Valve curation.” Even the hand-picked features have to be won. Valve said this about curated featuring in the same 2023 talk: “They are earned, and it’s competitive. They’re also based on a mix of sales and other customer signals.”
There is one more line from that talk worth holding onto, because it contradicts the way players usually talk about the store: “Steam never makes a permanent decision about your game. Every game has a chance to be successful.” The algorithm’s opinion is only ever about this week, and it carries no memory of the weeks a game lost. That is the mechanism underneath every game that got better after a rough launch: the store keeps reading this week’s signal, and last week’s has already expired.
What does not affect Steam visibility?
This is the part that surprises almost everyone, developers included. Valve’s own documentation rules out three metrics the internet obsesses over.
Wishlists are mostly not a factor. Valve states that “with a few exceptions like the Popular Upcoming tab, wishlists are not a factor in your game’s algorithmic visibility on Steam.” A million wishlists will not put a game on a store shelf it did not otherwise earn.
Store page traffic is not a factor. “The traffic landing on your store page is not a factor in your Steam store visibility.” A developer can buy a hundred thousand clicks and move nothing.
Review score is mostly not a factor. As long as a game sits at Mixed or above, meaning 40% positive or higher, the score does not drive visibility. It matters below that line and stops mattering above it.
Click-through rate belongs on the same pile, and the reason is a counterintuitive piece of arithmetic. CTR is page views divided by capsule impressions. Marketing analyst Chris Zukowski has argued that the number is not worth worrying about because it moves the wrong way: a game with few impressions, seen only by people the developer sent there, posts a flattering ratio. A game Steam is pushing across the store gets sprayed at millions of low-intent browsers, and its CTR collapses while its actual page views climb. A falling CTR is often the first sign that things are going well.
Laid side by side, the signals sort like this:
| Signal | Does it drive Steam visibility? |
|---|---|
| Buying and playing the game | Yes. Valve calls it a strong signal, with revenue as the primary reading |
| Supported languages | Yes. Steam is more likely to show a game to players who speak a language it supports |
| Wishlists | No, with a few exceptions like the Popular Upcoming tab |
| Store page traffic | No. Valve states traffic landing on your page is not a factor |
| Review score | No, as long as the game sits at Mixed or above, meaning 40% positive |
| Click-through rate | Not a signal Valve names, and the ratio falls as algorithmic exposure rises |
The first five rows are Valve’s own statements in the Steamworks documentation and the 2023 devcom talk. The click-through rate row is analysis, not a Valve position.
How does the Discovery Queue decide what to show you?
The Discovery Queue is the closest thing Steam has to a personalized feed, and it runs on a tag weighting system that is more specific than most players realize.
Steam knows how long you play everything, and it knows the median playtime for every tag across the entire user base. Your taste is inferred from where you are an outlier. Zukowski describes the logic with a clean example: if the typical Steam player puts one hour a week into games tagged Platformer and you put in twelve, the system can safely bet you like platformers. What it reads is the gap between your hours and everyone else’s. Your wishlist and your reviews do not come into it.
For top-selling games, the queue is typically the second or third largest source of store page traffic. Which games enter it, and at what threshold, Valve has never published.
One lever is unusually easy to underrate: language. Valve’s documentation is explicit that “Steam will be more likely to show your game to players that speak a language supported by your game,” so supporting more languages puts a game in front of audiences it was previously never shown to. A game can be effectively invisible to players who speak one language and healthy among speakers of another, for no reason other than which languages it supports.
So what do wishlists actually do?
Two things, and both are narrower than the reputation suggests.
First, the exception to Valve’s own rule: Popular Upcoming, the pre-release shelf, is the one exception Valve names by name, the surface where wishlist counts do feed the algorithm. Valve has never published a threshold, and it has never confirmed how the shelf is ordered. Unofficial community estimates cluster around 7,000 to 10,000 wishlists, with developer reports suggesting recent velocity matters alongside the raw total. Treat those numbers as folklore built from developer reports, not as a published rule.
Second, and more useful to you as a player: Update Visibility Rounds put a game’s major updates in front of “customers that have your game in their library or on their wishlist.” A wishlist is a subscription. It is how a small studio reaches you again a year later without paying anyone.
We wrote a full guide on how to build a Steam wishlist that works for you, and this is the mechanical reason it is worth the effort.
Why is half of Steam invisible?
Because the supply is enormous and player attention is not. SteamDB counted more than 19,000 games released on Steam in 2025, the highest annual total in the platform’s history. Nearly half of them finished the year with fewer than ten user reviews. Around 2,200 got no reviews at all. Roughly 7,100 peaked at nine or fewer. Only about 1,200 games, some 6% of the year’s releases, cleared 500 reviews.
This is not a new failure. In 2024 the same distribution held: 43.56% of releases sat under ten reviews and 7.2% crossed 500.
Read those numbers next to Valve’s design and the picture resolves. The algorithm follows player interest, so a game with no players generates no signal, so it gets no surface, so it gets no players. Obscurity is the default state, and it is self-reinforcing. The system never formed an opinion about quality in the first place.
That is also why the rigged-casino model of the store does not fit. Valve does not sell placement, does not sell store ads, and has said it keeps things that way deliberately. The store is not blocking good games. It is blind to games nobody has found yet, which is a different and more fixable problem.
What does this mean for you as a player?
Four things follow directly from the mechanics above.
- Playing is the strongest signal you can send. Time and money are what the algorithm reads. Reviews are for the humans.
- Your wishlist is a notification channel. It will not lift a game in the rankings, but it is how the developer reaches you when the game ships or gets a major update.
- Your Discovery Queue is a mirror of your playtime. If it keeps showing you the same genre, that is because you keep playing it. Playing something different is the only way to retrain it.
- The best game of the year might have eight reviews. With 2,200 releases at zero, deliberate searching finds things no feed will hand you. Our indie game iceberg and Steam hidden gems lists are both built for exactly that kind of digging.
Where our own game sits in all this
We are Overbaked Studio, a small independent studio in Istanbul, and our first game is Grandma Left Me a Building, a cozy apartment management sim where you restore the run-down building your grandmother left you and decide which of 30+ odd tenants, spread across four social tiers, end up living next to each other. It is unreleased, which means by the definitions above it currently generates zero of the signal Steam actually measures. No playtime, no revenue, no ranking.
What we have instead is the part the algorithm does not read: hand-drawn cartoon art in a capsule, a tag set that decides whose Discovery Queue we could ever appear in, no time pressure for people who want to plan at their own pace, and an opera singer who bursts into an aria in the elevator. None of it becomes a signal until someone is actually playing.
Every indie developer asks you for a wishlist. Here is what ours buys us: it puts our launch and our updates in front of you, and it counts toward Popular Upcoming, the one shelf Valve names where the number moves anything at all. It will not push us up a chart, and nothing else on this page will either until the game is in someone’s hands.
Frequently asked questions
Does wishlisting a game actually help the developer?
Yes, in one specific way. Valve’s documentation says wishlists are not an algorithmic visibility factor apart from a few exceptions like the Popular Upcoming tab. What a wishlist reliably does is subscribe you to that game’s launch and major update announcements, which is how small studios reach an audience they have already earned.
Why does my Discovery Queue keep showing me the same kind of game?
Because it is built from your playtime. Steam compares your hours per tag against the median across all players and treats the outliers as your taste. Playing a different genre for a few hours is what shifts it.
Can developers pay Steam for better placement?
No. Valve does not sell store ads or promotional placement, and has said this is a deliberate policy so the store does not become pay to win. Featuring is driven by sales and player signals, and there is no cheque that skips that step.
Why do good games disappear on Steam?
Usually because they never appeared. Over 19,000 games launched in 2025 and nearly half ended the year under ten reviews, so a game without early players produces no signal for an algorithm designed to follow player interest. Valve does say every game keeps its chance permanently, and a game can surface years after release if interest arrives.
Key takeaways
- Steam follows player interest. Valve’s stated approach is to react to what players buy and play right now, with revenue as the primary reading of that interest.
- Wishlists, page traffic and review score are not visibility factors, by Valve’s own documentation, other than a few exceptions like Popular Upcoming for wishlists and a floor at 40% positive for reviews.
- Click-through rate moves inversely to success. Wide algorithmic exposure lowers CTR while raising real traffic, so the ratio misleads in both directions.
- The Discovery Queue reads your playtime per tag against the global median and recommends from where you are an outlier.
- Invisibility is the default. Of 19,000+ releases in 2025, nearly half ended under ten reviews and only about 6% cleared 500.
If you want to send the one signal that actually counts for a game that has not launched yet, you can wishlist Grandma Left Me a Building on Steam.
