Valve has quietly ended more than a decade of hand-picked promotional deals on Steam, replacing the curated Daily Deal, Midweek Deal, and Weekend Deal calendar with a fully personalized recommendation algorithm slated to go live in early 2027 — and the first months of testing already showed that smaller games stood to gain the most from the shift. Valve announced the changes on September 29, describing the move as a step toward letting developers focus on making great games rather than competing for calendar slots.

An Era of Curated Gatekeeping Ends

For over a decade, landing a Daily Deal, Midweek Deal, or Weekend Deal on Steam’s Discounts & Events tab was among the most coveted outcomes in PC game publishing. Valve’s staff hand-selected a small roster of titles each week, giving them prominent placement in front of Steam’s 130-million-plus active user base alongside custom promotional artwork and steep discounts. The results were extraordinary for the titles that won slots — a Weekend Deal running Thursday through Monday could generate 100,000 units and 60,000 wishlists in four days for a single title, while a single Daily Deal typically drew over 10 million impressions during its 24-hour window, translating to revenue of anywhere from $10,000 to $2 million.

The mechanism was built on scarcity. Each day had at least two Daily Deal slots; each week had up to six Midweek and six Weekend Deal positions. Valve selected games based on an opaque assessment of existing sales momentum, review counts, and platform health — and curated slots were generally limited once per year per title, with at least a three-month gap between promotions. In 2024, across the full year, 2,843 games from 1,160 developers were featured in Daily Deals — a meaningful figure that nonetheless represented a tiny fraction of the tens of thousands of titles active on the platform.

That system has now ended. Valve announced on Monday, September 29, 2026 that the Discounts & Events section will move to fully personalized algorithmic recommendations in early 2027, and has already paused scheduling new curated placements. Developers who had existing deals booked will keep those slots. Everyone else enters a new era.

How Steam’s Recommendation Algorithm Actually Works

The change is not simply a policy shift — it is an architectural one, and understanding the architecture is what separates developers who will benefit from those who will not.

Steam’s recommendation system uses collaborative filtering techniques: a method that builds a model of each user’s preferences from their purchase history, wishlist activity, and the behavior patterns of users with similar taste profiles, then matches discounted games to the users most likely to buy them. This is qualitatively different from the old curated calendar. The old system took one featured game and showed it to every user who visited the Discounts & Events section — 130 million potential viewers served the same recommendation regardless of interest. The new system serves different games to different users simultaneously: the 3 million historical strategy fans on the platform see a discounted grand strategy title; the 2 million fans of cozy simulation games see a different one.

This is the source of Valve’s reported 10x figure. During the test period conducted over the past few months — in which Valve replaced some curated slots with dynamic personalized recommendations — the company reported surfacing 10 times more games to customers per day in that section. That is not 10x more games per user; it is 10x more distinct titles served across the full user base, because each user now sees a different selection tuned to their own history. The result was “meaningful increases in store page visits, wishlists, and cart additions,” per Valve’s own announcement.

The practical consequence for developers is immediate: success under the new system is not about lobbying Valve for a calendar slot. It is about how clearly a game’s metadata — its top 20 Steam store tags — signals its genre, subgenre, and playstyle to the recommendation engine. Steam’s official tag documentation confirms it uses all 20 visible tags for generating recommendations, and that niche tags carry disproportionate weight. A game tagged “Party-Based RPG” will benefit more from that specific tag than a game tagged only “RPG” or “Action,” because fewer competing titles share it, making the match signal stronger. Valve’s May 2026 tag taxonomy overhaul — in which 17 new tags were added, 28 were removed, and several were merged — was directly preparatory for this shift, cleaning up the metadata layer the algorithm depends on.

The Numbers That Spoke Before the Announcement

Valve had been running the test quietly since mid-September 2026, and the results surfaced before any official communication. Simon Carless, the founder of GameDiscoverCo and among the most authoritative independent analysts of Steam’s commercial dynamics, was among the first to flag that the featured games in the Discounts & Events section appeared to have become more personalized — describing the shift as a welcome development for smaller developers. The data that followed was striking.

Tim Bender, CEO of publisher Hooded Horse, analyzed traffic to an unnamed indie title on the platform and shared his findings via LinkedIn. In the five months preceding the algorithmic shift, the game had received 14 visits from the Discounts & Events section. In the single month after the personalized recommendations went live, it received 7,341 — making the section its primary traffic source on Steam, a position it had never previously approached. Bender also noted a qualitatively different personal experience on the shopper side, describing how obscure catalog titles with fewer than 100 reviews were appearing in his feed because they matched his specific taste profile. “This change took a source of traffic that was non-existent for niche indies,” he wrote, “and suddenly made it the best source of traffic in the entire store.”

Valve’s own stated rationale aligns: “Our goal is to let developers focus on crafting the most compelling games and deals for customers, and not spend time trying to convince someone at Valve to feature their products.”

What the Algorithm Still Cannot Do

The opportunity is real. The limitation is equally real, and the draft coverage of this announcement largely ignores it.

Collaborative filtering requires behavioral data to function. A recommendation engine can only match a game to likely buyers if it has enough signal — purchases from users with identifiable taste profiles, wishlists added by users whose broader preferences are known, playtime data that establishes genre affinity. A game with no purchases, no wishlists, and no engagement history is invisible to the system not because Valve has excluded it but because the algorithm has nothing to work with.

This is called the cold-start discovery problem, and it is a structural limitation in every collaborative filtering system. The old curated system gated on sales and review thresholds — you needed an established commercial track record to qualify for a slot. The new system gates on behavioral-signal richness — you need accumulated wishlist data, purchase history from genre-aligned users, and tag metadata clean enough for the algorithm to categorize you accurately. A game that launches as a “shadow drop” with no prior marketing, no wishlist base, and no existing user-behavior signal will not benefit from this shift in any meaningful way. The structural barrier has moved; it has not been removed.

This has a specific implication for launch strategy. In 2025, over 19,000 games launched on Steam; approximately 9,327 — nearly half — received fewer than 10 user reviews, and around 2,229 received none at all. Those games had insufficient signals under the old curated system. Under the new system, they face the same structural problem with a different name. The developers who will benefit from Valve’s change are those who have already built a signal-rich presence on the platform: games with active wishlist campaigns, strong tag profiles, and an established audience that the collaborative filter can use as a matching base.

Where Shadow Drops Become Riskier

The implications extend to launch timing strategy. Under the curated system, a developer could apply for a Daily Deal at any point after launch, providing a second-chance visibility window regardless of launch momentum. Under the algorithmic system, visibility accumulates continuously and compounds: a game that builds a wishlist base pre-launch, runs an effective Steam Next Fest demo, and tags itself accurately has an expanding pool of behavioral data for the algorithm to work with. A game that launches without those signals starts with an empty data profile and has no guaranteed mechanism for recovery.

Industry analysts have tracked this dynamic: the Steam algorithm consistently produces what developers call “blessed” games — titles that accumulate enough early behavioral signals for the algorithm to start actively promoting them, which produces more signals, which produces more promotion. The counter-pattern is equally well-documented: games that land quietly at launch rarely see algorithmic momentum arrive later.

Steam Next Fest — the platform’s recurring demo showcase — remains one of the few mechanisms that can generate behavioral signals at scale for a game that has not yet launched. A strong Next Fest performance builds wishlists, playtime data, and user-behavior patterns that feed directly into the post-launch recommendation signal pool. For developers navigating the new system, Next Fest is now even more structurally significant than it was before.

What Stays the Same

The seasonal sales calendar is unaffected. Steam’s Autumn Sale begins October 1, and the Winter Sale is scheduled to run December 17 through January 4, 2027. Valve published its promotional event calendar through the Summer Sale ending in July 2027, which continues to include dozens of themed genre festivals — from Auto-Battler RPG Fest to Couch Co-Op Fest — that provide their own discovery windows, particularly for genre-specific titles.

Alongside the algorithmic shift, Valve also promised new and updated developer tools for managing discounts and sale events, though it has not yet published specifics. The full transition timing will be confirmed “in the upcoming weeks,” according to the announcement.

A Platform Betting on Personalization at Scale

There is a direct commercial logic behind Valve’s move. A recommendation system that simultaneously serves different relevant games to different users can, in theory, maximize the aggregate probability of a transaction across the entire user base — a far more efficient use of the platform’s promotional real estate than serving the same game to every visitor regardless of taste. At the scale of 130 million active accounts, the difference in conversion efficiency between a targeted recommendation and a universal one is enormous.

The long-tail distribution argument here is theoretically compelling — Chris Anderson’s foundational 2006 work argued that digital distribution with negligible stocking costs enables a distribution of demand where niche items collectively rival the hits, provided users can actually find them. The early data from Hooded Horse’s test period suggests that for mid-catalog indie titles with clear genre identities and established signal bases, that theory is now operating in practice on Steam. Whether it holds for the lowest-signal-density games — those with no existing audience to generate the collaborative filter’s matching data — is the open question that early 2027 will resolve.

For developers who spent years locked out of the curated spotlight, that question matters enormously. The algorithmic shift is genuinely the most significant structural change to Steam’s promotional architecture in years. But its benefits are not distributed uniformly. They flow to the games that have already done the groundwork the algorithm requires: the wishlists built, the tags optimized, the behavioral signals accumulated. Understanding that requirement is not optional; it is the actual content of what Valve announced on Monday.

Frequently Asked Questions

What exactly is changing in Steam’s Discounts & Events section?

Valve is replacing its manually curated Daily Deal, Midweek Deal, and Weekend Deal system with a fully personalized, algorithmic recommendation feed. Instead of Valve staff selecting a fixed roster of featured games each week, the algorithm will serve each user a different selection of discounted games based on their specific purchase history, wishlist data, and the behavior patterns of users with similar taste profiles. The full transition is scheduled for early 2027. New curated placement scheduling is already paused.

Why did the early test results show such a dramatic traffic increase for some indie games?

The 14-visits-to-7,341-visits increase reported by Hooded Horse CEO Tim Bender reflects a specific dynamic in how personalization changes who sees a game. Under the old system, that title competed for one of a small number of fixed curated slots that Valve awarded based on existing sales performance — which it did not have. Under the new system, the same game runs a discount, and the algorithm routes it to users whose behavioral history suggests they would buy it. The game’s total potential audience hasn’t changed; what changed is the platform’s ability to identify and route it to the right subset of users. For games with clear genre identities and established behavioral signal bases, that difference is significant.

Does this mean indie games with few reviews will now get discovered automatically?

Not automatically. The recommendation algorithm depends on collaborative filtering — a method that requires existing behavioral data (wishlists, purchases, playtime) to match games to likely buyers. A game with no wishlist base, no prior purchases, and no established user-behavior signal suffers from what computer scientists call the cold-start problem: the algorithm has nothing to work with. The shift helps games that have already built a signal-rich presence — strong tag profiles, an active wishlist campaign, prior engagement. It does not create visibility for games that have generated no signals at all. The structural barrier to discovery has shifted, not disappeared.

What should developers do differently in the new system?

Three priorities become more important under algorithmic ranking. First, optimize Steam store tags: make sure your top 20 tags are accurate, specific, and ordered by relevance. Niche tags carry more algorithmic weight than broad category tags because fewer competing games share them. Second, build your wishlist base before launch — Steam Next Fest demo performance generates behavioral signals that feed directly into post-launch recommendations. Third, avoid shadow-dropping a game with no prior marketing; a launch with no existing behavioral signal is harder to recover from under collaborative filtering than under the old system, where a developer could apply for a curated slot months after release.