Cool!
It is very useful.
With it, I can finally accurately analyze the reasons for the fluctuations in home recommendations.
1 Like
I recommend monitoring your QPTR, as it trends similarly to UUQPTR.
You can still find “Unique user with impressions” and “Unique users with qualified plays” in Explore mode (click “Explore” in the top right corner of any chart and use the metric dropdown).
My game is based on fun, not robux, so can it get recommended? Please don’t take this post down, this is a question.
1 Like
It’s odd how our game released a better thumbnail, attracting more users, keeping a personalized thumbnail QPTR of around 1.80% yet in the last 7 days it has dropped to 1.35% after being stable for weeks. What is the point of that? I prefer our game to be displayed to users who would actually be interested in playing it in the first place, I assume it’s now being shown to users that don’t play games in the genre and that is causing QPTR to fall. Another one of our games has the same issue but a small drop at the moment, after stable QPTR for a long time.
I also wonder if it has anything to do with not having the current personalized thumbnail as your main thumbnail on the experience detail page, or it’s the icon being different that’s the issue. It’s a shame as our game had a large boost from the thumbnail and after a month any improvement is being erased. It would be nice to have some stability in the acquisition rather than these constant changes as our game stats have remained the same.
1 Like
Ever since Roblox started touting the “Similar Players” discovery algorithm our game has been thrown into the garbage in terms of popularity.
Our game Murder Mittens used to compete in the Survival category (this is a Murder Mystery type game), which our session time and retention were easily hitting the upper 50 percentile.
Ever since the change, suddenly we were competing against completely different game genres like roleplay games and fishing games, such as Warrior Cats: Ultimate Edition, Dandy’s World, and Be a Silly Seal - which pushed the target session time range our game “should be” far out of the water. We were also getting predominantly recommended to players who weren’t interested in the murder genre at all which ultimately led to people being curious; playing; and dipping immediately seeing they’d rather play a RP, story or fishing game - which also hurt our retention statistic.
We held an average of over 1K+ CCU last month, but now we’re barely struggling to even hit above 300.
And after this announcement, suddenly our players dropped even further. We had a low of under 100 lately and we’re hovering around 200 even at peak time. It seems that Roblox is reinforcing this new algorithm to affect developers even further.
It is genuinely disheartening to see that a disproportionate amount of developers, me and @Spiraxy included, are negatively impacted by the changes far harder than the benefits it provided to other developers.
I do not support this change.
10 Likes
What about if single player game is matching all signals excluding “7 day intentional co-play days per user” because it’s single player (no one can possibly join). This last signal will be ignored by algorithm for recommendations and will be counted as 5/5 signals? Also if game does not have any kind of Robux spending “7 day Robux spent per user” will be ignored too?
If a game has good engagement captured from other signals, recommendation algoritm does not ignore the game for home recommendations because of one signal is lower (example: 7 day intentional co-play days per user for a single player game, as referenced). We are looking to capture all types of games at different stages of development including engagement, monetization and coplaying game mechanics with these variety of signals in ranking. Ideal game can be strong on all those metrics, but it’s not necessary to get discovered on Home.
A fun game will have users playing the game and also playing more frequently. This is captured by 7 day playtime per user and 7 day play days per user. It can get recommended and you do not have to have all the metrics strong and positive to get a game recommended. Also, be aware users wanting to spend robux is a good indicator that they love the game (as it’s fun) and users are not only open to spending time playing but also open to spending robux. Keep building fun and engaging game!
Benchmarks displayed on analytics dashboard do not impact the recommendations algorithm but are meant to give you a point of comparison. qPTR, 7 day playtime per user, 7 day play days per user, 7 day spend days per user, 7 day Robux spent per user and 7 day Robux spent per user are the key factors into the recommendation algorithm for Home. Keep a close eye on D1 and D7 retention as the also tell how the metrics used for ranking are impacted. If users are not retaining, it can negatively impact is on playtime, play days, spend days, and intentional coplay days which further impact recommendations. Your game does not have to have all strong metrics, although all these metrics are strong for a highly retentive and engaging game.
Again, the benchmark displayed here do not impact the recommendations and only for a point of comparison.
1 Like
I feel this update, when i starting to play puzzle scape experience then when i leave, im staring to get similar experience and i found my self playing again and when i change my genre and decided to play simulators, i slowly getting more simulators.
Its like it depends on my moods, and the lists is also following that mood. I like it 
1 Like
if your game receives 0 home recommendation impressions (i.e. Roblox does not even try to show anyone the game, during time periods that all other statistics are healthy (some well above average, some average), what is driving this?
It seems like there is a purgatory aspect that a game cannot climb out of at some point for home recommendations. i.e. 11+ minute avg engagement from new users but zero home recommendation impressions, when occassionally any home rec impressions are even given, a few times there is a QPT rate of 100% (!), but no additional home rec impressions given after that. It’s not clear how any game could recover out of such a scenario.
I asked some people to play my game, and they loved it! They donated me hundreds so I’m making a donation system for rich people to donate if they liked the game. Also, thanks for clarifying it really made my day. I thought I couldn’t get recommended.
Currently users friending each other in a game is not counted towards 7 day intentional co-play days per user. Friends may join the experience in future and can count towards intentional co-play days. Not that we are continually evaluating what signals and inputs to our algorithm are valuable to tell us what makes users love a game and enjoy playing so we can give more discovery through recommendations. Focus on building retentive and engaging experiences, and we will share if we add additional signals.
1 Like
To be considered engaged, users generally need to spend a few minutes actively participating in your experience. This helps us differentiate between brief visits and sustained interest.
Benchmarks do not impact the recommendations algorithm but are meant to give you a point of comparison.
Users generally need to spend a few minutes actively participating in your experience to be considered engaged. This helps us differentiate between brief visit and sustained interest.
I love the algorithm changes and its great for it to be clear what goes into home recommendations.
However, adding onto @ph33sh. I have had several experiences where I had all stats in the 70th percentile or higher during sponsors and reached 200k+ visits via the sponsor system. And the algorithm never starts giving it impressions (ex. it might give 1000 impressions total). On the flip side I’ve had several games with below average stats and they’ve done exceptionally well in the algorithm (reaching millions of players), so its quite puzzling.
It’d be helpful if someone could explain whether I’m doing something wrong or if its a bug with the old home recommendation system. If it was a bug, is it resolved with this new recommendation system?
Sorry for tagging 
@BurningEmbar
What exactly does “actively participating” mean in this context? What does a player need to do in a game to be considered as a actively participating player by roblox?
This is awful for many developers. Not every game is meant to be played several times a week. Just as not every movie has to be re-watched. People still enjoy the movie, regardless.
As someone who produces tycoons, I can see how this ruins a lot of potential for story games and tycoons because these are not very replayable categories. Being un-replayable doesn’t make a game bad.
All this does is funnel players to the top experiences, really ruins all diversity on the recommended page. I don’t see how this will positively impact roblox monthly users.
5 Likes
Experiences will receive the same recommendation boost whether a user plays for 60 minutes or longer each day. We’re encouraging a variety of experiences and preventing any need to incentivize excessively long sessions. Playtime beyond 60 minutes won’t give experiences any additional discovery benefits.