News Data Deep Dive: How are new releases on Steam performing? [Steam blog]

Mor

Me llamo Willy y no hice la mili, pero vendo Chili
Sep 7, 2018
7,029
25,777
113


So, as it's usual, the Steam team has shared a blogpost with some figures about the performance of the games released last year in the platform and seems to be a good sign that tools (from Labs, new algorithms, etc...) might be working better than ever so this is helping games to reach more potential audiences, let's see what the team has shared and then we can discuss it. shall we?

We are constantly examining how the growth of Steam affects new titles and what it means for developers. Usually, our own internal questions mirror those of many in the development community.

In an effort to share some of our findings, we've put together the following post on how new releases have been performing. We're hoping this information is useful to developers, and we're interested in hearing whether more data-driven reports like these would be helpful to your product planning and development. And, if yes, what other topics would be of value. Read on, and please let us know what you think here in the Steamworks discussions.

The TL;DR

Here's a summary of what we found:
  • Over the years, there's been a continuous increase in the number of games achieving success on Steam. We'll dig into this in the first section below.
  • Earnings prospects for most - but not all - games improved in 2019. We'll dig into this in the second section below.

Some Background

A bit of history and context is helpful for understanding the analysis that follows. Prior to 2012, games that released on Steam were hand-picked by employees here at Valve. We realized that we were probably getting in the way of success for a lot of innovative games, and players were demanding more diversity in the kinds of games available via Steam. So, in 2012, we officially launched Steam Greenlight to allow players to vote on the games that would be released. Initially, we were only able to accept a small number of top-voted games per month. Then, in August 2013, our launch capacity improved with the introduction of some new internal tools, and for every month thereafter we accepted batches of at least 100 games. In June 2017, we launched Steam Direct with the intention of making the process of bringing a game to Steam more streamlined, transparent, and accessible.

Opening the platform resulted in a large number of new titles releasing on Steam, reflecting a diversity of niches and players we couldn't have dreamed of. Creating robust discoverability tools and systems was, and still is, crucial to ensuring that games will be surfaced to the customers who will want to play them, and we are continually experimenting with improvements to these systems.

So, has all of this been working? For whom? Read on for details.

More new releases than ever are finding success

One of our goals is to be a platform where great games can realize success. So a natural place to start this analysis was to ask: Are more games finding success? Of course, "success" is different for each developer, so we looked at several different benchmarks of success in this analysis.

Regardless of how we defined success, though, we found that an increasing number of games were achieving it.

As a first pass, we tallied the number of games earning over USD$10,000 in the first two weeks after release each year. For reference, most recent games earning around $10,000 in the first two weeks earned between $20,000 and $60,000 over the course of 12 months following release. We looked at the first two weeks of earnings to give even weight to games released across the year, and because initial earnings are a key metric that we hear many developers using in conversation. We also decided to study "paid games" (games with an up-front price tag) and save free-to-play games for another study, since initial earnings probably aren't the right metric for the success of free-to-play games. (For more on the reasoning behind our analysis choices, plus additional charts that use alternate success metrics, see our research appendix.)



As the graph shows, the number of games meeting this success measure has increased pretty consistently over time, with an 18% increase in 2019 compared to 2018. That increase wasn't just due to a larger number of games on the platform - the proportion of games meeting success increased by 11% in 2019.

You may also notice a jump on the graph between 2013 and 2014. That reflects the increase in the number of games we accepted to Steam beginning in August 2013. Many of these games weren't immediately ready for release, so the effect of that increase only really shows up in 2014. By 2019, more than three times as many new releases met the $10k benchmark than in 2013.



Many recent successes would never have previously been on the platform

As mentioned above, we opened the platform because we believed that "hand curating" the titles released on Steam had led to some great titles being missed. To test this, let's suppose we had never opened the platform. The dashed line in the graph below estimates what the trend would have looked like if we hadn't increased the quantity of games accepted to the platform in mid-2013. Assuming the trend was stable, it suggests that opening the platform more than doubled the number of releases meeting the $10K benchmark each year. While we can't say for sure, we think the green portions of the bars above the dashed line are, largely, games that would never have previously found success on Steam... because they never would have been released on the platform at all.



The $10k threshold was a fairly arbitrary starting point, and to make sure this analysis wasn't a fluke, we also tested higher and lower cutoffs and different time ranges. We found similar patterns in all cases. For example, we see more than 4 times the number of games earning over $5,000 in the first two weeks in 2019 vs 2013, and more than 3 times the number of releases earning over $250,000. You can find graphs for these and other benchmarks in our research appendix.

Most games did better in 2019 vs 2018

We were encouraged by the results above - there are more hits than ever these days, and it looks like our decision to open the platform helped that happen. But we also wanted to study how releases across the distribution – such as the median release, the 25th percentile release, and the 75th percentile release – were doing. (The 'median release' is the game that half the releases did better than and half the releases did worse than. You could also call it the '50th percentile release.')

How we studied median and percentile growth

When running these numbers, we were initially tempted to compare the median and percentile results from 2019 to all previous years. However, we noticed that the median game in 2019 was a very different kind of game than the median game in, say, 2013. And the same was true for the 25th percentile game. We realized the composition of games on Steam had changed so much between different release policies that a full historical comparison was meaningless. (See our research appendix for more on how we reached that conclusion.)

Instead, we compared 2019 to 2018, isolating the comparison to the first full year of Steam Direct. This comparison allowed us to more accurately see how the market on Steam is settling into the new policy. Additionally, we made several changes to our discoverability systems in 2019 in the hopes of better matching games to customers who would want them, and this comparison gives us some insight into whether those changes are working.

First, we looked at changes in median earnings. We found that the median game released in 2019 earned 24% more during its first two weeks of sales than the median release in 2018.



Next, we looked at percentiles other than the median. Here, the news was more mixed. On the upside, the 75th percentile release (meaning the release earning more than 75% of new releases in each year, but less than 25% of new releases) earned 56% more in its first two weeks in 2019 vs. 2018. However, the 25th percentile release earned 17% less.

More generally, we found that releases above the 35th percentile earned more money in 2019 vs. 2018, and releases below the 35th percentile earned less.



We want to make sure we aren't standing in the way of games' successes, so we plan to take a closer look at everything that contributed to these results.

What's next

We hope this study was helpful and gives more insight on the impacts of Steam's growth on new releases, both recently and over the years. We know there's more work to be done to help developers find success on Steam, and there's a lot more to be studied.

We’d like to get your input on what other analyses might help you better understand the landscape. Let us know your thoughts and suggestions here in the Steamworks discussions.

Note: If you're interested in a look at our methodology and some more detailed graphs, please see our research appendix.

-The Steam Team

Now, as we can see more and more games seem to have benefited from the tool changes compared to previous years and while official figures will always lack from some points (as Valve is not allowed to share figures from third partners due to contracts and such) it is great to see that the changes they did are working well.

What do you think about this changes and the results? let's talk about them.
 
OP
Mor

Mor

Me llamo Willy y no hice la mili, pero vendo Chili
Sep 7, 2018
7,029
25,777
113
Well, I appreciate the figures shared and this gives us a glimpse of the general picture that it's really interesting (they even shared an historical chart in their index page, no, not the headset haha) but still would like to know a few more things, for example, what's the performance of the generally called "shovelware" or more specific numbers, let's be realistic, this last thing will never happen due to the contractual obligations between Valve and third party developers/publishers out there who don't want their data shared for their own reasons, still would love to see this happening even for once.

Now, in terms of performance, it is clear that the tools that have been appearing on Steam since last year thanks to Labs have helped a lot to the smaller devs, sure, not all will reach success but that's an impossible scenario based on the market movements, and let's be clear, those who get bad results it's because they use their storefront in a bad way looking actively for those kind of games, for example, the hentai puzzle games, if you curate your own store by putting those on the ignore list, the system won't recommend you that kind of product that often (I made my own personal test with more than 30 daily discovery queue and recommendations were pretty on point most of the time even if the product itself didn't appeal me personally)

Anyway, I appreciate a lot their data, sure I want more but this still gives us some light.
 

eonden

MetaMember
Dec 20, 2018
275
930
93
I think Ars Technica article is also quite interesting:

Also, I will steal a post from other place:

The algorithm is "stealing" views from lower performance games into higher performance (and there is a point at higher percentiles were the views would be rather superfluous so there is not that big of a growht on 90+) which is why you will always have a dip on the lower performing games (sadly).
Expanding on the "stealing" views. There is of course a case that the views are stolen because the conversion between view -> sale is much better in the games that are being promoted, so it wont be a 1:1 conversion of decreased earning on games -> increased earnings due to algorithm but rather a 2:1 or something like that.

If it were possible to see the natural grow of Steam in revenue between 2018 and 2019, we could lower the curve by that % and see which percentiles beat the "Steam growth" mark and that would point towards what game benefited the most by the algorithm (likely the bell shape part) and what percentiles didnt benefit from the algorithm nor the Steam audience growth (probably until percentil 50?). Kind like this (the line I put is just a random line).



The views are "stolen" from the orange and yellow part (as they do not beat the natural Steam growth) and given to the "green" part. That is still somewhat flawed because part of the natural steam growth is also probably due to the algorithm (so some of the top yellow games would also be benefitted by it) and as I commented before, not all audiences grow at a same pace.
The Yellow part, even if it saw an increase in earning, still didnt beat the natural steam growth (thing of it as "Steam Market Inflation") so their "net earnings" actually saw a decrease.


Of course, my idea of a general "Steam Growth" is flawed because Steam is not one single audience but many audiences that grow at a different pace, so it is not really possible to compare the revenue growth of games that are "roguelite" with "hardcore fps". Still, you should never use "noise" in general as a first point solution.
 

Durante

I <3 Pixels
Oct 21, 2018
3,835
18,379
113
Of course, my idea of a general "Steam Growth" is flawed because Steam is not one single audience but many audiences that grow at a different pace, so it is not really possible to compare the revenue growth of games that are "roguelite" with "hardcore fps". Still, you should never use "noise" in general as a first point solution.
Great post overall, but this is a particularly good point.
I feel like you can see this "many audiences" theme directly when really impactful games happen. E.g. the last one was probably PUBG, and it completely changed the economics of "should I provide a Chinese translation" on Steam (and many publishers are still catching up).
 

eonden

MetaMember
Dec 20, 2018
275
930
93
Great post overall, but this is a particularly good point.
I feel like you can see this "many audiences" theme directly when really impactful games happen. E.g. the last one was probably PUBG, and it completely changed the economics of "should I provide a Chinese translation" on Steam (and many publishers are still catching up).
Yeah, as I said, I would really like if Steam provided similar information but to some of the different core "tags"/"genres" as a way to show devs what kind of things have had a bigger improvement.

Over language, it would be interesting to see the effect of adding or not different languages on the same population. Although of course, the problem of that is that games that have different languages would often have a higher budget and thus be more biased towards higher percentile games. I guess Valve in general has done a decent job of pushing the Chinese language as an important thing (but I guess devs would need to have more accurate data)..

Still the information is quite interesting and much more clear than most other platform holders give lol.

Regarding the (sadly) lower revenue games, it is still quite interesting that most of them actually saw improvements on revenue (as 80% of 2019 games were under the 5k in 2 weeks threshold and only the lower 35% saw their revenues decreased in general).
 
  • Like
Reactions: lashman