User avatar
Site Admin
Roblox discovery signals: what the Recommended for You algorithm actually ranks
Roblox discovery signals: what the Recommended for You algorithm actually ranks

If you have ever wondered why some Roblox games get thousands of players while others struggle to reach double digits, the answer often comes down to the Recommended for You algorithm. This post breaks down what signals the algorithm uses, how the explore and expand phases work, and what you can do to improve your game's visibility.

How the recommendation system works

The Recommended for You sort on the Home page uses a two-stage process:
  • Retrieval: the algorithm selects a subset of games that a user might enjoy based on engagement, retention, and monetization signals.
  • Ranking: the algorithm takes the retrieval results and ranks them in a personalized order for each user.
Signals from sponsored ads, curation, search, charts, friends, teleport, notifications, and social media sharing can accelerate your game's consideration for organic discovery. However, once a user arrives at your game through Recommended for You, only the engagement and retention of that organic cohort determines whether the algorithm continues to recommend your game.

Key signals that affect ranking

The algorithm uses many signals, but these carry the most weight:
  • Play through rate: the rate at which users play your game after seeing it in Recommended for You.
  • First play bounce rate: the rate at which users leave after a short session. This is a negative signal. High bounce rates suggest players are not finding enough to stay engaged early.
  • Play days per user: the average number of unique days users engage with your game.
  • Playtime per user: the average time users spend in your game, capped at 60 minutes per user per day.
  • Intentional co-play days per user: the average number of unique days users come back to play with friends.
  • Qualified play sessions per user: meaningful play sessions that filter out accidental clicks or quick bounces.
  • Spend days per user and Robux spent per user: monetization signals.
The explore and expand phases

Roblox uses an explore and expand model. When your game gets a spike in new users from recommendations (explore), the algorithm watches how that cohort engages. If they stick around and spend time in your game, Roblox is more likely to recommend your game to more user cohorts (expand).

This means a single content update that brings in engaged players can have a compounding effect on your visibility.

What hurts your visibility
  • Leading with giveaways: metadata that implies monetary rewards is not prioritized for recommendations.
  • Mismatched metadata: if your title, thumbnail, and description do not match the actual gameplay, the algorithm will not recommend your game.
  • Non-unique games: if your title and visuals closely resemble an existing game, you will rank lower.
How to improve your discovery
  • Keep your metadata accurate and unique. Use original imagery and naming that you or your team created.
  • Focus on retention and engagement. Improving these directly enhances your recommendation signals.
  • Use experience events to keep your community engaged and drive traffic to your game.
  • Maintain your Roblox Group with social links and regular updates.
  • Avoid renaming your game too often, as this reduces the chances players can find it.
Sources
[1] Roblox Creator Documentation: Discovery
[2] Roblox Creator Documentation: Promotion