The home page is recommending absolute bottom of the barrel games. The algorithm is being played by these games to appear more frequently. It is taking up space where other more relevant games could be shown.
I don’t expect 19% like ratio games and a game called “Women tycoon” to be shown to me on my home page just because they spam tags.
The recommendation and search algorithms should be smart enough to completely filter out these experiences from ANY recommendations. I thought tag abuse was fixed years ago. If it wasn’t, please consider filtering out any games with excessive use of tags.
I’m interested in a staff response on this as it could be many things.
Of course, no algorithm is perfect; it caters to people differently based on many factors. A factor could be that you play games with a similar like rating, player count, genre, general terms, etc.
Also, yes, the spamming of tags is said to be something to avoid:
Avoid spamming – Avoid repeating keywords or adding irrelevant keywords. This may result in demotion of your experience.
Again, no system is perfect, so any keyworld spam prevention could catch things that are fine. And with no way as a creator to see those issues ourselves, we’re left in the dark. I guess this is a feature request kind of thing, but I’d like to see a warning if Roblox flags the game for spam or something else that hinders discovery.
I’m all for a better system to prevent spam. I am worried that I might get caught up in it though, despite not spamming words in the desc.
Anyways, interested to see a response. Maybe it is just a bug, or multiple bugs.
Currently, we are actively working on the balance between encouraging creativity within a single type of experience (e.g. racing games), and reducing spam and clickbait in each user’s recommendations.
As we continue to fine-tune our algorithm, we are prioritising reducing the prevalence of clones and spammy content in your recommendations. However, we expect that high-quality experiences that ride popular trends might still be shown to you.
In the examples you provided, I understand that the like ratios are low and they appear to have a large number of tags.
We expect to release a post soon that will contain more information on our strategy towards combating spammy/clickbait content, and best practices for developers to avoid being mistaken for spam.
