GameRecommenderMatch

Discover your next favourite game in Steam. Based on your public profile information, last played games, hours played across different games, categories and even your wishlisted games, we try to offer the most probable match for your next purchase. Use filters to get even better results.

How does the Steam game recommendation tool work

What is GameRecommenderMatch?

If you love games like we do, you probably wasted countless hours looking for the next game that will hook you. The purpose of the GameRecommenderMatch recommendation tool is to build an algorithm that allows gamers to find the best game without wasting that time, and with much accuracy as possible.

For now the tool is starting and the recommendation algorithm might not be perfect, but with the help of all the users we can build the perfect tool that reduces the time we spend searching for a game, and directly log into it.

What data do we use for game recommendations?

GameRecommenderMatch uses your public profile data; we do not require any kind of login. If your profile is private the tool will not work since it cannot retrieve your games, hours played, etc.

Are all the Steam games shown in recommendations?

No. This recommendation tool excludes standalone game demos, DLCs and games tagged as NSFW. This is simply to grant that the content shown to every user it's safe. As for the demos and DLCs we will be adding them in future tool updates.

How often is the data updated for game recommendations?

Maintaining a 24/7 updated DB costs money. For that reason, right now the multiple databases are updated weekly. This does not affect prices since that data it's retrieved on the fly at the time you use the tool.

Do you use AI for recommending games?

The recommendation tool uses your own profile data, crossed with other players like you. It uses playtime, games, game tags, and it updates based on the different available filters.

What does the tool actually weigh to calculate a match?

Three things, in this order of importance:

  • Taste match (the largest share). How closely a game's tags and genres line up with what you already play. Your own games are not counted equally: hours played count on a curve, so 200 hours weighs more than 20 but not ten times more; anything under an hour is ignored, and the 1-2 hour range is discounted heavily because that is usually card farming rather than playing. Games you played recently count for more than old ones, on a gradual decay, and something you have touched in the last two weeks gets an extra push. Finishing achievements in a game is read as a sign you genuinely liked it, and playing far longer than the typical player counts for more still.
  • Players like you. What people with libraries similar to yours own and play, which is how the tool suggests things your tags alone would never surface. This is where the "because you played X" notes come from.
  • Quality (the smallest share). The game's Steam review score, adjusted for how many reviews it has, so a game with 50 positive reviews does not outrank one with 50,000.

On top of that: games from a studio you have real hours with get a bonus, bigger when the game also matches your tags and bigger for the developer than for the publisher - a familiar name alone will not push a weak match into your list. Something already on your wishlist gets a small nudge to break ties. Your filters are applied on top of all of it, and a few deliberate wildcards are mixed into the results so the list does not become the same safe suggestions every time.