Interactive recommender doesn't work properly

[Dead Miside]

It says it forms suggestions based on my library and playtime, using some kind of "machine learning".

So I tried it and it just recommended me the same ♥♥♥♥ I get to see on the main steam page. Like DOTA 2, WARTHUNDER, DOKI DOKI, VR CHAT, etc. I never play MOBA's, I never play low-quality f2p crap from the CIS, I never play japanese-styled games, nor VN's, all the less so japanese-styled VN's, and for sure I don't have VR.

So what's the point of this "steam labs"? It doesn't differ from default steam recommendations of ungodly crap you can get on the main page. Where is this "machine learning" part? It just ignores my library and playtime completely and shoves DOTA-like promotional dung straight into my face, literally the same algorithm as in the store. Is it supposed to be this way? Or am I doing something wrong?

[Crazy Tiger]

On top of that interactive recommender is a tab called "discussion". Please post it in there, as that's the correct place for both feedback and questions.

[aiusepsi]

If you want suggestions for more niche games, then drag the "popular - niche" slider all the way across to "niche".

I'm not sure what you're expecting to see from hearing that it uses machine learning. In this case, it means they trained neural net using play data (what games you play, how long and when, probably) from all Steam users, and then that neural net makes predictions about what it thinks you will like based on your play data. It doesn't take stuff like tags or if a game is F2P or VR or whatever into account explicitly at all.

If you're getting a lot of stuff like Dota 2, etc. it just means that people with similar play histories to yours probably also play a decent amount of Dota 2, etc.

[Dead Miside]

Originally posted by aiusepsi:

If you want suggestions for more niche games, then drag the "popular - niche" slider all the way across to "niche".

I'm not sure what you're expecting to see from hearing that it uses machine learning. In this case, it means they trained neural net using play data (what games you play, how long and when, probably) from all Steam users, and then that neural net makes predictions about what it thinks you will like based on your play data. It doesn't take stuff like tags or if a game is F2P or VR or whatever into account explicitly at all.

If you're getting a lot of stuff like Dota 2, etc. it just means that people with similar play histories to yours probably also play a decent amount of Dota 2, etc.

Well, I tried that "niche" slider, and apperantly for steam analysts the word "niche" is synonymous with "indie memeware straight outta trashcan greenlight", so it's a "pick your poison" situation.

As for that machine learning part, I find it hard to believe that people with the most of their playtime spent in Killing Floor and Europa Universalis love playing games like DOTA 2, since DOTA 2 is literally #1 suggestion for about a HALF of that popular-niche slider. The same goes for those warthunder and crossout, which seem more like a sneaky advertisement instead of grounded suggestion. I don't believe any self-respecting person would play those two games unless they are paid to advertise it (which happens too often on youtube and twich), so it's a very suspicious to always see them as top recommendations, when my library has literally nothing to do with their kind.

Thanks for replying though.