Here comes the algorithm - Valve updating Steam to show different games for discounts and events

I thought they were bringing dynamic pricing when I read the title and had a heart attack 😅

Youtube is saying that they are doing this too and it lead to small players basically being utterly excluded as established big channels simply generate more engagement.

Lets hope Valve is doing a better job, but I kinda hate the allmighty algorithm.

I hope they make it easy to block stuff, because algorithms feeding on engagement can also be fed on negative engagement (fb's specialty) or brigaded (twitter, to get things to trend or to bury stuff).

And that they take responsibility for Steam's startup pop-up by keeping it manually curated.

Countdown until someone tries to sue and regulate Valve as an advertiser because they got an algorithm now. lol

Kinda on the two sides with that. In one hand its really nice to improve the algorithm to show you the deals you most likely care about, but on the other hand i always feel like it will push people more into their own bubbles with recommended games that they already have played sort of. Nothing new and different will get discovered easily. At least that's what happens in Youtube.

The issue with how they do the curated deals is that right now everyone has to fight for visibility. Games only get a couple of guaranteed feature slots for their entire lifetimes, and the rest relies on the whims of Valve.

This documentation page explains how the Curated features currently work:
(External Link: You need to be logged in to view this.)

Nice! I too don't generally bother looking at those discounts and events because they're mostly "games that lots of people like so we think you'll like them". I, unfortunately, am not lots of people and have my own inscrutable eclectic tastes – I don't care how many times Steam shows me Skyrim Release #428, Balder's Gate 3, or The Witcher 3, I'm just not particularly interested (and if that ever changes, I know where to find them). I picked up Dressmaker (External Link: You need to be logged in to view this.) which released just last week, but only because I randomly stumbled upon it outside of Steam. I'm looking forward to getting more personally tailored (pun not intended…) recommendations from Steam in future.

Some of the best games I've tried, and the same is true with music and films, are from recommendations from people that did not play the type of games I do. I believe there is a lot of value to play stuff that I would not usually do, and expand knowledge, fun, experiences, etc. Leaving an algorithm to do this it's probably not in our best interests.

Quoting: ArehandoroSome of the best games I've tried, and the same is true with music and films, are from recommendations from people that did not play the type of games I do. I believe there is a lot of value to play stuff that I would not usually do, and expand knowledge, fun, experiences, etc. Leaving an algorithm to do this it's probably not in our best interests.

Valve's featured deals are largely unhelpful for this goal, too. It's mostly the same AAA games you'd see anywhere or ultra popular "friendslop" stuff recently, it's nothing you wouldn't find by looking at social media trends with extremely rare exceptions.

I'm willing to give them the benefit of the doubt because the release calendar they put on the front page, which is also user customized, has been really good at showing me interesting things and also established stuff I would already want to see together, it has been a really good balance for me and put a good amount of games in my radar.

Quoting: MblackwellThe issue with how they do the curated deals is that right now everyone has to fight for visibility. Games only get a couple of guaranteed feature slots for their entire lifetimes, and the rest relies on the whims of Valve.

This documentation page explains how the Curated features currently work:
(External Link: You need to be logged in to view this.)

The pitfall of the next system: Generic AAA-games which are tailored to "Everybody likes them!' will most likely always take the top spot due to them generating the most buzz, leading the algorithm to think that this is something everybody is interested in.

It kills niches, because they are small by nature. Butbwhen the algorithm now tries to judge your interest, it doesn't see mechanics. It sees features and similar played games. And then the most popular amonst them will get recommended first.

Because why should the algorithm recommend an unpopular game barely anybody cares about? If so few people care about it, player X will most likely not care either! So let's dumb it!

That's the problem YT has. It doesn't matter for the platform holder. They make sales. But it makes it incredibly hard for the small fish in the pond.

I might be wrong and for some reason Valve can solve this problem. But I won't keep my hopes up.

Quoting: Savor592That's the problem YT has. It doesn't matter for the platform holder. They make sales. But it makes it incredibly hard for the small fish in the pond.

I might be wrong and for some reason Valve can solve this problem. But I won't keep my hopes up.

I feel for Valve here, it isn't easy to thread the Discovery/Recommendation needle when there's 500 new games every week. Every game dev wants their game featured for visibility, preferably all of the time; many players don't want Tiny-Indie-With-500-Reviews recommended, they want known AAA quantities; others want tiny indie games shown, but not THAT one; meanwhile few want to actually spend time curating their feed, so giving those options just sounds like work to a lot of players.
We somehow expect Valve to magically divine our inner desires in games, and unless they nail it 100% of the time, they're doing it wrong.

Quoting: Savor592Because why should the algorithm recommend an unpopular game barely anybody cares about?

Because the financial incentives are for Valve to have it do so if that game fits a particular user's style. The whole point is to tailor recommendations to people based on the kinds of things they like (using proxies like tags and playtime on the games they buy and play) rather than just "what's most popular". If I play a lot of relatively niche games centered around certain themes and mechanics and show no interest in the most popular stuff, then it's in Valve's financial interests to show me more of those kinds of games because there's a higher chance I'll find something new that I like and want to buy, rather than showing me EA SPORTS FC™ 27 for the umpteenth time.

Quoting: Philadelphus

Quoting: Savor592Because why should the algorithm recommend an unpopular game barely anybody cares about?

Because the financial incentives are for Valve to have it do so if that game fits a particular user's style. The whole point is to tailor recommendations to people based on the kinds of things they like (using proxies like tags and playtime on the games they buy and play) rather than just "what's most popular". If I play a lot of relatively niche games centered around certain themes and mechanics and show no interest in the most popular stuff, then it's in Valve's financial interests to show me more of those kinds of games because there's a higher chance I'll find something new that I like and want to buy, rather than showing me EA SPORTS FC™ 27 for the umpteenth time.

I suspect that the current learning phase process still applies. If the game doesn't get large amount sales right at launch, it takes some time for Steam to learn who to recommend the game to. Half baked asset flip probably doesn't really deserve any sales, but the good stuff needs to be seen by more people. Steam just can't know which one it is if the developer is the only one who has played the game.

Quoting: Philadelphus

Quoting: Savor592Because why should the algorithm recommend an unpopular game barely anybody cares about?

Because the financial incentives are for Valve to have it do so if that game fits a particular user's style. The whole point is to tailor recommendations to people based on the kinds of things they like (using proxies like tags and playtime on the games they buy and play) rather than just "what's most popular". If I play a lot of relatively niche games centered around certain themes and mechanics and show no interest in the most popular stuff, then it's in Valve's financial interests to show me more of those kinds of games because there's a higher chance I'll find something new that I like and want to buy, rather than showing me EA SPORTS FC™ 27 for the umpteenth time.

That's the problem: Valve makes sales regardless. And with a sub 10 review indie game the algorithm will have way less certainty that showing this game will result in a sale. The certainty will be high for a AAA game people have shown to buy regardless of their genre preferences.

Last edited by Savor592 on 1 Oct 2026 at 10:00 pm UTC

Quoting: druciferCountdown until someone tries to sue and regulate Valve as an advertiser because they got an algorithm now. lol

They always had.
Even within the popular usage of the term.
Even within the common marketing definition of the term.

When the press and popular culture nowadays talk about algorithms they talk about recommendation algorithms.
Valve had one and they have one.
The old one just lacked a certain form of complexity the current one doesn't.
They trained their people to put certain games on the recommended page and certain games not.
They had a personalized feed and thus even some level of automation.

The common marketing definition of the term excludes transparent algorithms.
Valve's recommendation algorithm was both complex and proprietary. It was totally opaque.

Quoting: Savor592That's the problem: Valve makes sales regardless. And with a sub 10 review indie game the algorithm will have way less certainty that showing this game will result in a sale. The certainty will be high for a AAA game people have shown to buy regardless of their genre preferences.

They apparently consider this a risk worth taking, and given that Valve is a business rather than a charity I'm inclined to consider that they have good reason to believe this change will ultimately help their bottom line or they wouldn't be making it. (That's not to say it can't also help indie devs get more visibility and sales and people find games they like but wouldn't have otherwise found, it could be a win-win-win situation.)

Valve has data beyond simply "is this game popular or not" by which they could personalize recommendations; they then just need whatever algorithm they make to not include popularity when determining games to recommend. From a computer science standpoint it's not hard; from a business standpoint, they seem to think this will help their bottom line. It's not hard to imagine how, either, because people are not actually all sheep that automatically buy AAA games regardless of their genre preferences simply because they're popular, or the current situation where everyone gets shown the same curated list of popular games would be the best one for Valve's financial interests and they wouldn't be moving away from it.

Do we think this change has happened already? I just received an email from steam saying the Autumn sale is on, and it says it has "5 recommendations for you" ... which are

1- Grand Theft Auto 5 enhanced
2- Heroes of Might and Magic Olden Era
3- Palworld
4- Civ 7
5- Arc Raiders

That list is WILDLY out of context for my gaming tastes and Steam itself knows that from the billion hrs I've played in all the genres I play. I do not give one shit about gta, palworld, civ, or arcraiders, and only MILDLY give a shit about a homm game. wtf happened here?

The announcement says "Improvements coming in early 2027", so presumably it hasn't changed yet.

Although, later on it says they have been experimenting with it:

Our goal with Steam’s discovery features is to show the right game to the right customer, so we spent the last few months experimenting with replacing some curated spots with dynamic and personalized recommendations. By featuring games in a more personalized way in the Discounts & Events section, we were able to show 10x more games to customers in a given day in that section. As a result, we saw meaningful increases in store page visits, wishlists, and cart additions.

I HAVE to have been included in those experiments because that list they sent me is sooooooooooooooooo stupid. I mean it's like antisales. It makes me not even want to visit the store at all for anything if that's what they suggest for me. LOL.

Wouldn't that mean you weren't in the experiment and were instead getting the current system, where everyone gets the same "popular because it's popular" stuff regardless of personal preference? 🙂 The point of the change (which was only mentioned to have been trialed in the store so far) sounds like it's to recommend more stuff you would be (more likely to be) interested in, regardless of popularity. (Not that it'll ever be perfect, of course.)