In Dubit's design philosophy, a new Roblox game has about 30 seconds to win a player over. So we evolve a genre that is already popular, which cuts what a player has to learn. Roblox reports that since 15 June 2026 its Recommended For You algorithm has listed first play bounce rate among its "Most Important" signals. It is a negative signal: the rate at which players leave a game after a short play session. On 15 June 2026 Roblox also expanded the algorithm's view of retention from 7 days to 28.
New games take off on a trend wave. Dubit's read is that the biggest new Roblox games typically succeed by riding one, and sometimes more than one. One kind is internet culture, the memes and formats players already quote, such as "Brainrot" humour. Another is the genre itself. Escape games, wave survival, obbies and tycoons surge regularly, and once one is hot players go looking for new variations of it. The last is hype, an unexpected twist, social chaos or high-stakes play that gets a game talked about and streamed. Designing into one or more of them raises a game's chance of working.
We evolve a popular genre instead of inventing one. Teaching a player new mechanics takes time a new game does not have. We call it "evolution beats revolution". Building an original game inside a trending format lets it ride demand that already exists, and stand apart on theme, mechanics and progression.
Roblox groups its signals in two tiers. In Dubit's reading, Roblox's discovery algorithm since 2025 can drive hundreds of millions of plays to one game, and sustained scale comes from designing around its signals. Roblox's developer forum post of 15 June 2026 sets out the signals behind Recommended For You, the sort on its Home page, in two priority groups. It lists four as "Most Important". Play-through rate is the rate at which users play a game after seeing it in Recommended For You. First play bounce rate comes in two time windows, under 60 seconds and 61 to 180 seconds. The other two are play days per user and playtime per user.
Roblox lists four more as "Important": intentional co-play days, qualified play sessions, spend days and Robux spent, each per user. The other six signals are each averaged over day 1, days 2 to 7 and days 8 to 28. Every signal counts only players who joined organically from Recommended For You. Roblox's newsroom post of 15 June 2026 says it found in testing that overvaluing short-term engagement could disproportionately favour games that win attention with exciting thumbnails but do not deliver long-term value for players.
Did the Roblox discovery algorithm change, and what happened?
Yes. On 15 June 2026 Roblox expanded its Recommended For You algorithm from a 7-day view to a 28-day view that directly measures long-term retention. Roblox's newsroom post that day reports that, in testing, the updated algorithm did a significantly better job of surfacing games that retain players over time. Roblox says this led to an increase in daily active users and engagement across the platform.
Roblox's developer forum post of the same day says it removed qualified play-through rate as a signal and added play-through rate and first play bounce rates. Roblox's stated aim is to better understand what happens immediately after a player decides to click on a game. Roblox's newsroom post also says creators whose games create long-term player value, through retention, co-play and spend, will have more opportunity to build a lasting audience.
