Valve explains “How Steam decides which games are recommended”



When you access the game distribution platform ``

Steam ,'' you will see a series of currently popular games, as well as recommended games based on the games you have purchased. Eric Peterson, in charge of development at Valve, which operates Steam, explained how these games are displayed to users. The video below also has Japanese subtitles, making it very easy to watch.

Steam Visibility: How Games Get Surfaced to Players - YouTube


According to Peterson, Steam's main goals are to 'match player preferences with games' and 'build long-term relationships with both players and developers.' A key focus of these efforts is to gain confidence from players that the games displayed on Steam are ``relevant to them'' or ``match their interests.'' It is said that there is.

There are several rules that Valve uses to achieve these goals. One is that we don't sell advertising, and if you look at any page on Steam, it's obvious that there are no advertising spaces. Instead, they are trying to display content that will interest players even without ads. According to Peterson, the games that are displayed as recommendations are made up of a combination of algorithms and curation (selection by humans).

When you look at the Steam home page during sales, you may see games displayed in large size at the top. In this example, ' DAVE THE DIVER ' is featured, but this display frame is the most powerful marketing place on Steam, and the conditions for posting are very strict, and it is said that the displayed game will attract many users. It is said that it will be posted for the first time after giving strong confidence to Valve. The games displayed in this pane are determined by 'curation' and all users will see the same ones.



The 'Featured & Recommended' section is usually displayed at the top. This frame is called the main capsule, and different games are displayed depending on the user. The games displayed here are determined by an ``algorithm'', and for example, they may be recommended based on the ``tag'' of the game the player likes.



'Curation' is basically common to all users, and the displayed games change every day or every few days. On the other hand, the 'algorithm' displays different games depending on the user, and may not display games that the user owns or games that have tags that are set to be hidden.

The algorithm takes into account each user's play history and the games on their wishlist, as well as the games their friends are playing, the games their friends recommend, the games their friends have on their wishlist, and the region they live in. The games that will be displayed will be automatically determined based on the top selling games.

Below “Featured & Recommended” is the “Special” section. Games in this section are curated based on factors such as pre-scheduling and what games users around the world are currently interested in. There are many sales related items such as Midweek Deal and Weekend Deal, and all users will see the same items.



In order to be featured in promotions such as Midweek and Weekend Deals, a game needs to be a bestseller within a few hundred ranks on Steam, and it also needs to be recognized by Valve as a game suitable for promotion based on sales and user trends. That's what he said. If you want to be featured in these promotions, it's important to localize your game, set prices by region, and include marketing assets to widely promote your game.

Below that is the 'Discovery Queue'. The Discovery Queue is a recommended game summary frame consisting of ``new releases, top sellers, and titles similar to titles you've played before,'' and the games displayed are determined by an algorithm.



Below the discovery queue is the 'Recommendations based on the games you play' section, and by clicking on details you can search for games by setting conditions such as 'popular' and 'niche.' It seems that the games displayed are also determined by an algorithm.



Further below, you may see 'Curator Recommendations.' This section will not appear unless you follow at least one

curator . A curator is ``an individual or group who gives recommendations so that users can find games they might be interested in from within the Steam catalog'', and for Japan, they compile Japanese information for various games. There are communities such as `` Japanese Information '' and `` Unidentified Game Weather '' run by game commentator Moyashishamo.



In addition, there are various sections such as ``Recent Updates,'' ``Popular VR Games,'' and ``New Hot New Items.'' Mr. Peterson explained in a little more depth about the games selected in these sections.

First, I would like to explain about the 'New Hot New Items' section. This section automatically displays 'new releases' at the top of the list that meet certain criteria after they've been released.



Next is the 'Top Sales' section, which apparently decides which games will be displayed based on the total revenue for the past 24 hours, including DLC and soundtracks. The games displayed in these sections vary by region, for example in Germany you'll find

Baldur's Gate 3 and Counter-Strike: Global Offensive (later Counter-Strike 2).



On the other hand, in Japan, '

ARMORED CORE VI FIRES OF RUBICON ' and ' Apex Legends ' are at the top.



The Popular Upcoming section lists upcoming games that have the most wishlists. If a game meets the criteria, it will be eligible for display and will be displayed in order of release date. It seems that the trend of the wish list for the past two weeks is mainly reflected.



Lastly is the 'Special' section. Here you will find works that are ``discounted'' and ``popular.''



Next, Peterson mentioned some rumors surrounding Steam.

First of all, Mr. Peterson dismissed the rumor that ``a game's exposure is determined by the number of accesses to the Steam store page'' as being groundless and that such a mechanism does not exist. However, the higher the number of accesses, the more likely the game will be purchased, so it is definitely beneficial.



In addition, the theory that ``the degree of exposure is determined by the review score'' is also denied, with some exceptions. The exception is games with less than 40% positive reviews. According to Peterson, a game with a rating below 40% is 'mostly negative,' and games with such low ratings are rarely recommended.

Another thing is the wishlist. Just because a lot of users have added the game to their wishlist doesn't mean it will be displayed more often, but the above-mentioned 'Popular Coming Soon' is an exception because it refers to the number of wishlists. The number of wishlists does not affect whether an item is displayed on other pages, such as the home page or discovery queue.



On the other hand, localization has a big impact on recommendations. Steam games are recommended based on the region where the user lives, but if a game does not provide a language tailored to the region, the probability of being recommended is greatly reduced. Conversely, Mr. Peterson said that such localization is highly recommended because the more languages it supports, the more likely it will be recommended. In addition to localizing the game itself, localizing the display language of the store page is also an effective method.



Finally, Mr. Peterson gave advice to developers, saying, ``It is important to use the Discord community and other SNS for marketing rather than just publishing the store page.If you can reach a large extent with the first step, then The reach of the game will snowball.There have been cases in the past where the game became a big hit simply because of its quality, but this is extremely rare.The key is to raise awareness and build momentum as much as possible before release. ” he said.



in Web Service,   Video,   Game, Posted by log1p_kr