11 votes

Introducing Steam Labs - a place for testing and giving feedback on experimental Steam features

8 comments

  1. [3]
    nothis Link
    I tried the Interactive Recommender. Seems about time Steam fed all their data into a machine learning algorithm to improve their recommendations. It's okay, but it shows me games I already own...

    I tried the Interactive Recommender. Seems about time Steam fed all their data into a machine learning algorithm to improve their recommendations. It's okay, but it shows me games I already own (?) and still seems to be stuck with "you liked Celeste, so you probably like every single 8bit indie platformer game under the sun" type recommendations. It's almost as if less personalized and more quality-oriented curation would have a better chance of pointing me towards interesting games.

    Like, I don't care about genres, really, I care about quality. Despite the 20000 games released a year, there's only like 50 quality indie games a year on Steam, still, how about you "filter" for that and I might discover something new? Like, I played some Football manager game religiously for a short while in the late 90s and some deer hunting sim in 2005 or so. Those are as absurd genre choices for me as they're probably for most people, here, but it was fun to discover that stuff. With more and more "personalized" algorithms, you're starting to find yourself in these bubbles where you're constantly recommended a dozen shitty games of some random genre you recently played.

    7 votes
    1. [2]
      Whom (edited ) Link Parent
      Yeah, I've never seen automatic recommendations for any kind of art be all that useful. I dunno, maybe it works for the kind of person who finds out they like Star Trek and as a result they go...

      Yeah, I've never seen automatic recommendations for any kind of art be all that useful. I dunno, maybe it works for the kind of person who finds out they like Star Trek and as a result they go down the path of finding and watching all sci fi with similar characteristics, but it seems really shallow to me to find new things by what they are.

      Is something good? Does it do things that are fresh? Does it say something which is interesting to me? Do I learn something new from it? Will it stick around in my head? Those are things that matter to me when deciding if I want to play something, and they're things which algorithmic suggestions might be able to scratch at but which I've never seen actually happen, and certainly won't on Steam where tags beyond just genres and technical things like "multiplayer" or "controller support" are very rare (though I do appreciate some like the "Masterpiece" tag) and don't seem to be used for much of anything outside of directly sorting by them.

      This recommender seems better than most, but it still remains that if I want to find games just like Skullgirls for some reason, I can just search "2d fighters" and be done with it. Does anyone really need help finding more games of a certain type? Things like this (and the million similar things for music or whatever else) seem like they're tackling problems that the existence of search engines already solved much better a long time ago.

      4 votes
      1. nothis Link Parent
        I'm at a point where my understanding of good "game recommendations" is almost entirely incompatible with search and automation. I want to "discover" stuff. Something new, something I haven't...

        I'm at a point where my understanding of good "game recommendations" is almost entirely incompatible with search and automation. I want to "discover" stuff. Something new, something I haven't played before. But it's also gotta be good? It's almost defined by not being possible to deduct from looking at what I played before, like I want to be surprised.

        Now, of course, 90% of people (arguably the people that bring in the money) don't care about "discovering" games. They're happily provided for just by linking to the top 20 best selling games on Steam. But most people who actually bother to look for stuff have probably a similar interest in discovering something new. Even if you get burned here and there, it's worth it because you get to play a more varied selection of games.

        Ironically, this is where we might have gone full circle: It's probably a full time job to look at games and analyze them for how interesting they are. It's the job of a game critic. Not an averaged number based on a thousand critics, no, one game critic who can actually make connections and point out specific nuances in the games they recommend. So, indeed, Steam might have a more interesting front page by employing 10 (not a thousand, ten) people who have some taste in games and who have a human-beings-like discussion about what to put on the front page. Beneath the top-20, of course, but still. It's not an impossible job. For the 20,000 games release a year, 95%+ is probably trash that can be filtered out at a glance. The rest is probably easily identifiable by keeping an eye out for prerelease buzz on gamedev forums and Twitter and whatnot.

        Basically, let anything on, who cares, but give the job of choosing games to human beings.

        5 votes
  2. [3]
    Deimos Link
    Direct links to the Labs page: https://store.steampowered.com/labs There's also a separate blog post for the "Interactive Recommender" experiment specifically:...

    Direct links to the Labs page: https://store.steampowered.com/labs

    There's also a separate blog post for the "Interactive Recommender" experiment specifically: https://steamcommunity.com/games/593110/announcements/detail/1612767708821405787

    6 votes
    1. [2]
      TheJorro Link Parent
      This recommendation thing.... actually works?! I played with it for a bit just to see what I don't already have in my large (~900) library and I found about a dozen games within a few minutes. All...

      This recommendation thing.... actually works?! I played with it for a bit just to see what I don't already have in my large (~900) library and I found about a dozen games within a few minutes. All interesting looking stuff too!

      If this is just the start, I can't wait to see how it performs in a year.

      4 votes
      1. moocow1452 Link Parent
        Would be nice if they had one for games you already owned but haven't played or only tried out.

        Would be nice if they had one for games you already owned but haven't played or only tried out.

        2 votes
  3. Whom (edited ) Link
    The microtrailers twitter account was a brilliant idea, Valve is very smart to pull from that. I hope it takes off! Imo most of the problems people have with Valve not curating their platform...

    The microtrailers twitter account was a brilliant idea, Valve is very smart to pull from that. I hope it takes off!

    Imo most of the problems people have with Valve not curating their platform (aside from the hate speech issues) come down to Steam being very poor to comb through and quickly get information from. ...and thinly-veiled disdain for the "true" amateur indies, but that's beside the point. :P When using Steam, you're either you're assaulted with games you already know about or you're in a sea of $1 game logos and can't tell anything about what you're looking at. This should do a lot of good to open it up, though it would be best if they integrated this stuff into their search and sort results. I want to see these while I sort through point and click adventures on sale or games under 50 cents or whatever!

    5 votes
  4. ThyMrMan Link
    I honestly think I like the automatic show more than anything else. It just feels simple and easy to put up on my second screen and glace up if something interesting shows up. Only major issue I...

    I honestly think I like the automatic show more than anything else. It just feels simple and easy to put up on my second screen and glace up if something interesting shows up. Only major issue I have is I feel like each clip needs to be a bit longer, or better clips get picked that show off more actual gameplay.

    3 votes