Post by redonihunter in Build and indie game discovery and suggestion platform

Viewing post in Build and indie game discovery and suggestion platform

Well, from my impressions, the projects "announced" here, are most often made by developers for developers to get a vehicle for promotion. The pitch is not for people to find games, but for devs to promote games. (Your's is a better approach).

I found the discussion I mentioned, it might be of interest to you /t/4748289/itchio-api-for-game-metadata and /t/5258962/games-recommendation-system-dev-progress

Steam tags are interesting, because they do have a ranking. With enough popularity, they should become more accurate. Then again, if one curator runs in the wrong direction in the beginning, a lot of people might just agree and not question the wrong classification. This is also bound to be influenced by trends over the years. Glaring examples: https://thejaspel.itch.io/backpack-hero on Steam the game has no indie tag. Or the fact that peopl voted "rpg" the 15th most important tag for https://store.steampowered.com/app/646570/Slay_the_Spire/ . That game is not even close to a rpg, but opinions obviously differ here.

What it should do is surface DD1 to the player who logs “dread, strategic, investment” and DD2 to the player who logs “fast, replayable, casual runs” and let the player decide.

The thing is, dd2 is not replayable and not fast and not casual either. A run will last like 3 hours. And you do not replay it, you grind it. And that grinding means, it is not casual. Replayable means, you get worth out of a new replay. Maybe you associate this with rogue-lite, because neither dd1 nor dd2 is tagged with "replay value". Or you just used things you think are opposite of investment.



	Darkest Dungeon	Darkest Dungeon II
1	Turn-Based Combat	Turn-Based Tactics
2	Dungeon Crawler	Roguelite
3	Dark Fantasy	Dark Fantasy
4	Roguelike	Dungeon Crawler
5	RPG	Lovecraftian
6	Lovecraftian	Singleplayer
7	Difficult	Roguelike
8	Turn-Based	Perma Death
9	Singleplayer	Turn-Based
10	Perma Death	Turn-Based Combat
11	Strategy	Party-Based RPG
12	Atmospheric	Gothic
13	Dark	Story Rich
14	Indie	CRPG
15	2D	2D
16	Roguelite	RPG
17	Horror	Hand-drawn
18	Gothic	Turn-Based Strategy
19	Fantasy	Procedural Generation
20	Great Soundtrack	Stylized

I see here that tag usage changed over the years. Even on Steam with their limited tag pool. What's with turn based tactics not even appearing in dd1? New tag? And what's with all the rpg tags. Those games ain't rpg. And there is no perma death in dd2, hence it not being a roguelike anymore. And notice how it is no longer tagged indie. An interesting signal is the appearnce of story rich and hand-drawn. dd1 was hand-drawn too.

And I notice that some tags are more accurate than others. Lovecraftian for example seems to be rather accurate. And rpg seems to be noise selected by people that do not know that just because you control "a character" does not make a game an rpg.

So what does this actually say? Tags are not reliable. Because people have different opinions when to apply them and how important they are. That goes for selecting the tags, and for creating any signal from tags to match to a player.

What I initially meant with the 30 years, there has been reasearch done. Companies use classifications and player profiles and obviously techniques to match those two. Which means two things: 1, you might dig up knowledge how to do this. The theory behind it, and ways that will not work or are established industry standard. And 2, re-inventing this from scratch in any way that can compete with existing solutions is gonna be hard.

Yeah, having a mapping of feelings a game will invoke and searching for those is well on paper. But this sounds like a thing that cannot be gathered automatically from existing data.

Like Reply

evangelos2536 days ago(+1)

Sorry again for the slow response/ I wanted to actually sit with this one rather than reply fast and let it go. Plus last days were crazy.

Thank you for this, genuinely. The Darkest Dungeon vs Darkest Dungeon II tag table is a great example and it lands the point hard that DD2 isn’t fast, isn’t casual, isn’t really “replayable” in the way that word usually gets used, and no amount of automatic tag-scraping would have caught that nuance. I also appreciate the links, and the honesty about the “dread/strategic/investment vs fast/replayable/casual” framing and you’re right that it was closer to “things I assumed were opposites” than an actual accurate read of DD2, and calling that out is fair.

The bigger point you’re making is that tags are noisy because people disagree on both what a tag means and how much it matters, and that this problem exists whether you’re picking the tags or trying to derive a signal from them is one I don’t think I can argue with. And the industry research angle is well taken too. So if 30 years of classification/matching work already exists, trying to reinvent that from scratch with scraped tags is a bad bet, and I’d rather know what’s already been tried and failed than repeat it blind.

This actually pushed me to dig more into it rather than just take the mood/tag approach at face value so thank you for that, even indirectly.

It also gave me an idea, though it’s a different angle than “classify the feeling correctly.” If the real problem is that no automatic system can honestly say “this game invokes dread” or “this one is casual” because that’s a subjective judgment, not a fact, then maybe the answer isn’t to try to nail that judgment algorithmically at all. Instead I can surface the actual differences between two games side by side (whatever’s genuinely knowable, pace, length, structure, mechanics, permadeath or not and etc.) and let the player read that and decide for themselves what it means to them, rather than the site pretending to have resolved something as personal as “does this feel like dread or does it feel like a casual run.”

It’s a good idea, but it needs real thought before I’d build it. Mainly figuring out what data is actually trustworthy enough to put in front of someone that way, versus what’s just tags with the same noise problem you described. But the core of it is don’t claim the verdict, show the differences, let the player decide on how it feels like the right instinct, and it’s one I want to explore properly rather than rush.

More broadly this whole exchange gave me both ideas and motivation to keep pushing on this. I’ve genuinely enjoyed the back and forth with people here. It’s kind of rare to get responses where people actually think it through before replying, and it shows in everything you’ve written.

Thanks a lot man! If you ever want to share more ideas as this develops, or just see how it grows over time, I’ve got a Discord happy to have you there if that’s easier than threads like this one.

Like Reply