STEAM? Offerings in 2024 Will Approach 100,000 Games

There will be triple the number of available games as compared to 2019

INTRO

I’ve been floating a theory that the game industry boom and bust cycle is affected by game publication rates. Specifically, my assertion in Game Development and Market Failures is that unrestrained investment in a monopolistically competitive market results in failures for both investors and game developers.

The TL;DR of that article is:

1. Investors see a chance for a big return and collectively invest at rates driven by the size of the expected return
2. The industry eventually creates games faster than players can buy and play them
3. Players have a fixed amount of money to spend and and do so across all games
4. Many game developers cannot fund operations on their slice of the widely dispersed revenue so they shut down

The model is simple to simulate. The boom and bust cycle emerges endogenously, as expected. But do we have data to support the thesis?

A KEY CONDITION

A key condition of the proposed model is that as players have more and more games to choose from — and divide their spending across — fewer and fewer individual games receive enough revenue to support the developer. The more games there are, the smaller the cut for each developer.

So what are the game publication rates? What do we know about the overall availability of games?

There have been articles commenting on STEAM? publication rates, like this one from GAMERANT , but that was only about 2023. A lot of other evidence we have is anecdotal, like my own STEAM? account.

But now I has all the STEAM? datas, so let’s explore (PC) game publication rates and see what they say.

Note: one driver for this work, and hence for trawling STEAM? data, was a commenter on my article who wasn’t convinced. This one’s for you Sean Christian Bellinger ! ?? Thank you for the motivation.

Thanks for reading! Subscribe for free to receive new posts and support my work.

METHODOLOGY

The data I am using for this investigation is the same data as the previous digs . To be specific, all STEAM? data presented here is for applications labeled as a game with a publication date between 2003-01-01 and 2024-06-30. The data were pulled from public STEAM? sources.

I also include data from two more sources for annual and monthly game publication counts: Steam Spy and SteamDB .

Steam Spy and SteamDB

These two sources also mine STEAM? public data. I’d originally started with these sites when looking for evidence on game publication rates but when they couldn’t provide the level of detail I wanted, at least not in an easy-to-query format, I wrote code to pull data for myself.

GOOD SCIENCE NOTES:

• SteamDB uses, to the best of my knowledge, the same methods as I do.
• Steam Spy differs in at least one way: it samples public user profiles, analogous to polling methods. Those player profiles are not available via an aggregated API — they require HTML scraping that I have not implemented (yet).

EXECUTIVE SUMMARY

This is a high-level description of conclusions. The details can be found below for those who want to dig in. Or who wants to check my work. ??

Here’s what the data show:

• The STEAM? game publication rate has increased every year since 2019.
• The STEAM? game publication rate is increasing nonlinearly — it is accelerating.
• This is true even for games blessed by STEAM?’s secret-sauce quality algorithm (games not tagged as Profile Features Limited).
• STEAM? is on track to have 100,000 games, of which 25,000 are STEAM? vetted, by the end of 2024.
• If publication rates stay on this path STEAM? will surpass 200,000 games total and 50,000 vetted games — a doubling — by the end of 2029.

Here is maybe the most mind boggling statistic:

STEAM? will have three times as many games available at the end of 2024 as at the end of 2019.

And now on to the details!

ANNUAL PUBLICATION RATES

Let's start by looking at the annual game publication rates from all three sources: Steam Spy, SteamDB, and mine. To differentiate, I’ll refer to my own data as RobinDB.??

The data we have for 2024 is only through 06/30/2024. The 2024 totals are estimated by scaling data by the factor 365/181 (06/30/2024 is the 181st day of the year) and are plotted with a ?? symbol of the appropriate color.

SteamDB segments games into PROFILE FEATURES LIMITED and regular — I’ll refer to those as UNLIMITED — games. The graph above combines both types. We will come back to the LIMITED/UNLIMITED segmentation later in this article.

Some observations:

• On an absolute scale the numbers are close for the years before 2013 or so, and then start to diverge a bit.
• Steam Spy is consistently lower than SteamDB in estimating the number of published games.
• Steam Spy has a significantly lower estimate for published games starting around 2020 then starts to increase again in 2023.
• RobinDB rates are below the other sources up until 2021, when the data show a steeper increase than shown by either SteamDB or Steam Spy.
• RobinDB tracks SteamDB more closely, probably because they both use the same data collection methodology, whereas Steam Spy uses a sampling method of publicly visible user profiles.

Perhaps most importantly:

• The annual game publication rate is increasing every year since 2019.
• Estimates for the number of STEAM? games to be published in 2024 are in the range of 15,000 to 20,000.

MONTHLY PUBLICATION RATES

The annual graph shows an apparent drop from 2018 to 2019. The monthly rates are much noisier, but let’s look at those rates to see if we can see more detail.

The monthly estimates from our three data sources show at least four distinct eras — regions of consistent publication rate behavior.

SCIENCE NOTE:

• That assertion is not a statistical argument regarding differences of slope, just visual inspection. However, a bit of detective work — aka reading internet sources — provides logical reasons for the differences.

STEAM? Greenlight

The STEAM? Greenlight program has been discussed a great deal and was formally started in August of 2012. Unfortunately, although the goal was to increase the rate at which games could get approved, there was still a bottleneck in place.

A year later, in August of 2013, some changes were made to remove the bottleneck. Hence the need for a Greenlight Anniversary marker — that’s when publication throughput actually improved.

STEAM? Direct

The next important event was in 2017, when Valve made a significant change to STEAM? by retiring Greenlight and introducing STEAM? Direct. The new system allows anyone to submit a game to the store without needing community approval and after paying a small fee. The program began by retroactively approving 3400 games waiting in the STEAM? Greenlight queue . That likely explains the immediate jump in publication rate.

SCIENCE NOTE:

• I do not have an explanation for the essentially flat game publishing rate between 2017 and the start of 2020. But theory crafting — aka, speculation — is that it could be because investors had yet to discover the appeal of the game industry.

COVID-19

On March 11, 2020, the World Health Organization characterized the outbreak of COVID-19 as a pandemic. In the United States, many people were forced into the work-from-home environment and, more importantly, social isolation. Players had more time to play and spend on games. Game developers apparently also immediately started publishing more games, as well.

TOTAL AVAILABLE GAMES

Since the start of COVID-19 game publication rates are increasing on an annual and even monthly basis. What does that mean for players?

STEAM? publishes games of all genres and sizes, from single-person game teams to multi-studio efforts. Games remain available until they can no longer function, for example, in the case of a live service that shuts down.

So how many games are available on STEAM? for players to acquire and play?

Once again, the estimated 2024 totals are marked with a ?? symbol.

The simple linear estimate for games published by the end of 2024 indicates that STEAM? will have from 90,000 to 100,000 games available for players to acquire and enjoy.

RATE INCREASES ARE NONLINEAR

Inspecting the data reveals an accelerating annual rate. Let’s fit a function to get a feel for that nonlinearity and what it means.

We don’t want to fit the entire curve because we know that publication rates have had at least four eras with rates that differ significantly, so let’s focus on where we are now — the post-COVID-19 era. And because we are interested in totals, let’s fit the accumulated counts, not the annual or monthly individual rates.

Once again, the estimated 2024 totals are marked with a ?? symbol.

This graph shows fits for each data source of 6 data points using a quadratic.

SCIENCE NOTES:

• I am not suggesting that the process underlying publication rates is quadratic. I am unaware of evidence for the actual underlying process at this time, but a quadratic is a common and simple nonlinear function.
• The fits are good, which is not surprising given there are only 6 data points and three parameters. However, each data point is an integration — literally an integral — of many data points, so noise has already been smoothed out. The result for fitted models is reduced squared error.
• TL;DR - Metrics show improved fit for the quadratic even when penalized for the additional degree of freedom.
• Because I’m lazy and didn’t want to write more code, I used an online curve fitting tool to compare a linear to a quadratic fit. I was specifically interested in metrics that include a penalty for overfitting. In that group, the tool includes original and corrected Akaike Information Criterion, (AIC and AICc), the Bayesian Information Criterion (BIC), and the adjusted R2 value (aR2). They all gave the same story.

Using the end-of-2024 estimate as a starting point, these fits imply a doubling of the total available games some time during 2029.

To state this clearly: we could have over 200,000 games available on STEAM? before the end of 2029.

SCIENCE NOTE:

• Extrapolation using a model that does not capture the process underlying the data is a BAD IDEA?. It could be entirely wrong. We are also extrapolating 5 years out and the market could change. For example, investors driving the creation of new studios may look elsewhere in their search for alpha .

IS IT JUST MORE BAD GAMES?

The advent of STEAM? Direct abolished various human created and judged quality criteria. Basically, it removed that gatekeeping — anyone could pay a fee and publish a game.

That change caused some to worry that there’d be a flood of junk games.

I do think some of that has happened. 2018 is also when STEAM? relaxed other restrictions and allowed adult-themed games. I have a different article for game ratings and mature content , but I can say that adult-themed games do not account for the accelerating publication rate we are seeing.

Interestingly, in 2018 the STEAM? service also began tagging games as Profile Features Limited.

What does Profile Features Limited mean? I’ve not found a STEAM? definition of the term, but the SteamDB site describes it this way :

Steam automatically enables profile customization features as games reach certain player and sales metrics that give confidence that a reasonable number of customers that are engaged with the game.

If you see the term Profile Features Limited on a game, it didn’t pass those hurdles (yet).

We can use the tag as a divider of games into two quality groups as vetted by STEAM?’s secret-sauce metric.

SteamDB segments games by the Profile Features Limited on the store pages — the tag is not available through the API. So, I went back and scraped store pages.

Let’s have look at the data grouping by games that pass — or don not pass — that metric.

This time we call out our estimated 2024 totals by highlighting the bars.

The SteamDB and RobinDB estimates are close, and I intend to explore these differences in another article.

So let’s break out the totals for UNLIMITED games and see what that progression looks like since 2018.

Once again, we call out our estimated 2024 totals by highlighting the bars.

From Wikipedia, regarding the transition to STEAM Direct :

In the process of transitioning from Greenlight to Direct, Valve mass-approved most of the 3,400 remaining games that were still in Greenlight, though the company noted that not all of these were at a state to be published.

Notice that 2018 had a lot of games that include, I assume, at least some of the 3400 games grandfathered into the UNLIMITED category. Then in 2019 we see a dip as those criteria are more diligently applied.

That said, we still see a continuous increase of UNLIMITED games each year, so let’s plot the accumulated total and fit a quadratic again.

Once again, the estimated 2024 totals are marked with a ?? symbol.

SCIENCE NOTE:

• The fits are stupid good, but again not surprising given that the 7 data points are already noise-smoothed via integration. Again, this is not an assertion about the underlying process.
• TL;DR - Metrics show improved fit for the quadratic over a linear fit, even when penalized for the additional degree of freedom.
• I used the same online curve fitting tool to compare the linear and quadratic fits. All statistics show the quadratic is better, even when including the penalty to adjust for over fitting.

Using the end-of-2024 estimate as a starting point, these fits imply a doubling of the total available UNLIMITED games by the end of 2029.

To state this clearly: we could have about 50,000 UNLIMITED games, games of vetted quality, available on STEAM? by the end of 2029.

OUTRO

A key condition for the game industry boom and bust model that I’ve proposed is that players have so many games to choose from that they spread their spending around. From the executive summary we found that:

• STEAM? is on track to have 100,000 games, of which 25,000 are STEAM? vetted, by the end of 2024.
• If publication rates stay on this path STEAM? will surpass 200,000 games total and 50,000 vetted games, a doubling, by the end of 2029.

And for emphasis:

STEAM? will have three times as many games available at the end of 2024 as at the end of 2019.

I will humbly submit that these data support the existence of that key condition.

Anecdotes are not data, but here’s a personal experience: my STEAM? library is currently over 1350 and counting, and I have another 900+ games spread across 8 other launchers.

I haven’t paid real money for all of them. Many were acquired for free from marketing giveaways — especially in the last few years — or are free-to-play games. Scads were very cheaply acquired through “bundles”. ??

Although my personal horde is over 2200 games, I still pick up free games every week because the big publishers are all trying to get us to be sticky in their marketplace. Granted, my behavior can be characterized as completionist, but I don’t think I’m all that different from the rest of the gamer community.

I haven’t come close to playing them all, and many not to a significant extent, but some have hundreds and even thousands of hours played. It’s clear that I could entertain myself for quite a while — years — with what I’ve already acquired.

These data alone, and especially personal anecdotes, don’t prove the theory but the findings are consistent with the proposal.

OK. I think this is where I say, “Thank you for attending my Ted talk.” But there is more data to examine so, as a good scientist, I’ll keep evaluating.

GAME RECOMMENDATION: I am currently over 60 hours in on Islands of Insight , which I picked up for free from a STEAM? one-day event. I highly recommend it if solving puzzles with relaxing background music combined with offline play or multiplayer without toxicity sounds enjoyable to you.

Thanks for reading! Subscribe for free to receive new posts and support my work.

RESOURCES

?? My previous article: Game Development and Market Failures

?? My previous article: STEAM? Data Archaeology

?? SteamDB description of Profile Features Limited tag on STEAM? games

?? My previous article: STEAM? Data - Video Game Ratings and Mature Content

?? Wikipedia article on Population dynamics

?? Game Developer article on anniversary of Steam Greenlight

?? SteamDB releases page

?? Steam Spy API documentation page

?? The Verge article about the STEAM? Greenlight program shutting down

?? Game Developer article about the STEAM? Greenlight bottleneck and changes at the anniversary

?? PC Gamer article about the history and evolution of publishing games on STEAM?

?? STEAM? page for Islands of Insight

?? Wikipedia page on STEAM? Direct

?? Wikipedia page on STEAM? Greenlight

?? Investopedia article on the economic/investment term, alpha .

?? Eurogamer article Microsoft admits "many" PlayStation exclusives "better quality" than those on Xbox

?? Microsoft’s response to UK's competition authority — the Competition and Markets Authority (CMA) regarding competition

?? MyCurveFit - an online curve fitting and analysis tool . It is marked as "[BETA]” but it’s pretty darn awesome, compared to doing it yourself in Python or other language.