In this case, how do we know that the algorithm change isn't correcting a problem? I'm sorry for the game developer in the article, but every time his game was being promoted, somebody else's wasn't. How do we know he wasn't being over-promoted with the old algorithm? Without seeing everything that Valve sees, can we really know?

My guess is that because those games are so inexpensive, they probably sell a ton of units irrespective of price and category, and if the system is optimized to show stuff that is moving a lot of units in a short period of time, it will bubble to the top.

The other day, the only good recommendation I've seen in a long time came up: Battle Brothers has a new DLC, even long after the devs said they were not going to make any. But that's the only one I can really think of.

I think most personalization features on marketplaces are kind of overrated. I care way, way, way more about sales than I care about 'personalization.' I am just about never going to buy something at full retail apart from daily essentials.

I think his problem more realistically is that his category is extremely weird. Adventure solitaire RPG? How would you even begin to SEO that product page in a way that didn't cause a ton of confusion?

Even something like Thronebreaker, which is basically a solo player card game RPG, does not try to distinguish itself on search. It promotes itself based on the strength of the Witcher brand and on Gwent.

I'm going to be honest, given this information I'm wondering how you know you hate them. Any individual recommendation is likely low-quality, but that's mostly thanks to Sturgeon's law - Steam certainly has a good enough grasp of what I'm looking for in a game to point me at the right genres.

You would be right of course, but thats not the point. It's not harming anybody, and nobody has to use such a feature.

If you want more specifics: You complained about VNs as being "anime" and then said that all anime was about high schoolers. I provided recommendations for high-quality VNs that were neither anime nor about high school.

Some recs in that direction, based on what's presumably a collection of western RPGs and FPS games: Baten Kaitos (gamecube), Most Final Fantasy games in the single digits, Golden Sun I and II (gameboy advance), Umineko (PC), Steins;Gate (PC).

I'm not arguing against algorithms here, there are just too many games on Steam to show everyone a raw list of products and let them fend for themselves.. You need some kind of automated curation.

But as a seller of products, you have to put all your faith in the marketplace's ability to put your products in front of the right people as much as possible, and it can be frustrating when you suddenly start seeing data that shows this is happening less, and you don't know why.

..it's a gatekeeper, and we have a complicated relationship with those guys.

A decade ago or so I was tasked with writing a recommendation engine for a "creators community" company I worked at, and the feedback we got after launching that feature was pretty much constant for as long as I was still working there.

Everyone wanted to know exactly how it worked (so they could min/max for the logic) and no one was happy with the results.

I'm pretty sure there are a lot of complaints about it, but as much as Valve tries to better the situation there's a significant problem: It is hard to match people with games they'll like based solely on broad metadata about both.

> every time his game was being promoted, somebody else's wasn't

This isn't necessarily true. Imagine a recommendation list of 4 items being shortened to a list of 3 items... Or 4 small promotional thumbnail links being redesigned into a single larger image link for a single game. That larger image isn't going get 4x the traffic + clicks just because it's 4x bigger.

But that isn't the case here. "More like this" list was always 12 games. Both before and after the October change.

Anecdotally, I've been seeing users complain about Steam's recommendation algorithm for a long time, on gaming forums like ResetERA.

2 guys and a mic, without Youtube, would just be a self-funded hobby to show your friends and family, maybe enter some indie film shows, maybe help get a job in advertising or local TV with your portfolio. With Youtube, they might get a million subscribers and make on the order of $10,000 per year from what would otherwise be just a hobby! And there are on the order of 10,000 such channels with 1 million subscribers.

Now, you can say what you like about how Youtube treats the other 97% of channels with far less viewership - but there's a lot fewer than 10,000 channels on mainstream TV.

It's the same in game development, mobile app stores, music, sports, and a lot of other fields. The marketplace is open to a lot of people who really want in to get their chance at the top. It's hard to earn a living wage from doing the work when you're competing with people who are willing to earn less than a living wage.

2 million apps / 10 years = 547 new apps every day for the last 10 years. And of course the number of apps coming out today is far greater than it was 10 years ago so it's actually more like 1000 apps a day right now.

Steam has the same issue. Steam was around 8k new games in 2017. That's 21 new games per day, 153 a week, 666 a month. No one has time for that many games.

What would Steam do to suddenly get all those games to make a salaries for every indie that put stuff up? There's really nothing they can do. There's just too many games. At most Steam could go back to restricting the market. That wouldn't raise anyone's odds, it would just make invisible all the people that tried and failed to get on the market.

It's the same story with books. Apart from rare mega hits like Harry Potter, the bread and butter of the paperback world is least common denominator stuff like Danielle Steel and Nora Roberts.

I don't think we can blame Valve here. This has happened across all forms of media. There is too much content and not enough time.

This actually isn't a revenue thing, its a platform expectation thing. There's a really good video essay (on YouTube, naturally) explaining the 'Manufactured Authenticity' for popular YouTube channels: https://www.youtube.com/watch?v=8FJEtCvb2Kw

But this stands to reason. Mainstream TV show viewing figures refer to the number of adults 18-49 in the United States watching the show live(-ish now that L+SD and C+7 rates are used to sell ads) including at least putatively watching the various 3 minute blocks of commercials in the middle. YouTube viewing figures show the number of worldwide plays by all people in or out of the demo, most not in the United States and in less valuable markets, many adblocking, those who don't adblock often skip ads, and the ads are either inline or pre-roll. And the content is shorter, so it's sort of unclear how an 11 minute video should compare to a 42 minute television series.

Do you have a source for that? because it seems to me like there's a lot of youtubers out there making pretty good money.

A popular TV show averages in the neighborhood of 15 million viewers these days.

ADDED: So the parent comment seems pretty close. Maybe $150K/year gross revenue for viewership in neighborhood of the most popular network programming. That's in the ballpark of barely supporting two guys and their gear.

[1] https://www.inc.com/minda-zetlin/even-youtube-stars-with-14-...

YouTube’s cut is reportedly 45% - I am skeptical that the traditional media industry gives such a revenue share to its creators.

I agree that apples-to-apples is hard but the bottom line is that very few people make a real living from YouTube views.

Hundreds of thousands of people will try to get in on it, causing both the income of existing creators to drop, and the average income to plummet.

Any market where there is a huge supply of people who want to produce the good is going to have this problem.

100k daily views, £17 - £250 estimated daily earnings.

Game of thrones has an estimated 10 million viewers per episode. So, scaling up youtube earnings: £1,700 - £25,000 per episode. The actual per episode budget was around £4 million per episode.

See the gap here?

Actually this exists, it's called Youtube premium and there are some channels that offer some content exclusive to the platform (as far as I'm aware)

I have no idea how it's doing but I also don't have any interest in trying it out, I'd much rather support specific content streams via patreon or other offerings.

What should that particular anti-pattern be called?

I also dropped Pandora, because I had/have too many streaming music options and couldn't justify it. I don't remember commercials for their ultra premium offering, maybe its because I've been a paid customer for so long?

I would guess there's a very, very long tail of channels that have a very small number of videos and essentially no views.

I picked an example of someone who does it full time.

It won't be monetised, so you can't really compare it to TV?

Do all film actors make $60M/year just because Dwayne Johnson does?

I hesitate to even call it a scale. These things all follow viral dynamics: if you haven't blown up with billions of views, you're very possibly spending more than you get back.

Pareto Curve (more or less) of Y values on index X: [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,2,2,2,2,3,10,50,500,7000,9000000]

Index 30 is "off the chart"

Now revenue: $$*index

The correlation is linear (basically). The top youtuber who is a kid who reviews toys, makes the majority of his 22mil from ads, same as you or I would with <100 views per video.