[Article] Raph (Theory of Fun) Kosters - The Cost of Games

Link -> https://www.raphkoster.com/2018/01/17/the-cost-of-games/

This article charts the rising cost of game development (see above graph)

Then goes into the declining price to player per MB of installed game vs development cost per MB.



So the content size of games is going up as the resolution, processing and graphical power has increased.

However the production costs have levelled out, showing that game content production has hit productivity limits. e.g. Game Engines reduce the programming load but make for a market where more art/asset/content creation is needed to compete.

And Raph has three potential forecasts.

1. In 10 years Free games become the normal.

2. Super high priced AAA games with astronomic development budgets.

3. System or Procedural games reduce the development costs.


Now the Author states his data set is limited and his predictions could be off but this data set shows these trends that probably most of you are aware of.

The question is how can game and engine developers break the cost per MB trend, whilst maintaining good income from players expecting more from less?

PS My thoughts on the issues:

• Moore's Law hardware limits could crimp or reduce the upward demand for more content e.g. limit processing/graphical power growth.
• AI Helper and or Procedural Automated content generation could boost content production and reduce costs.
• Someone develops the 'Oasis', a super sized MMO sandbox that players and developers can modify to make and play the games they want.

 

For a good insight into the industry try Raph Koster's - Industry Lifecycles (Casual Connect 2008) presentation slides.

https://www.raphkoster.com/games/presentations/industry-lifecycles/

 

I don't think most indies have to worry about it..

 

I worked with Raph on a couple of games. His insights are always an interesting read, but I wouldn’t worry about it too much, they are high level observations from a particular point of view, not really actionable.

 

Really not sure how useful cost per megabyte is as a valuable metric. Especially with number 3, why would systems driven games be the outcome of having high res lightmaps and uncompressed audio?

 

Are there any Indie/Casual game data sets that prove your points e.g. Indie/Casual/Mobile games that are not growing in size, production costs and decreasing in price?

Lightmaps are part of the problem, the more assets/content you need to create to make a current generation game the more time/money it will cost to produce.

To reduce this production cost you could move to a more procedural systemic way to generate your games. E.g. XCOM2's use of procedural level from chunks or No Man's Sky procedural universe.

The thing is modern game engines do not have procedural toolkits for content generation built in, the first game engine to make an easy to use configurable procedural content toolkit will probably gain an advantage in the market.

However built in procedural content goes against modern free game engines asset store business models.

What would happen if No Man's Sky turned their game into a game engine?

 

@Arowx

No doubt the biggest challenge at the moment is content development and whilst game engines have become epically powerful the art content pipeline hasn't evolved extensively in general.

Although that's not to say there aren't answers, like Houdini for e.g. which is being adopted by AAA studios and indie's alike to try and offset some of the burden.. I can tell you know even from initial investigation it will save me weeks if not more in pipe creation alone.

I've never believed 3D modelling was just artistic nature, there's a fair bit of technical skill and with new tools the paradigms are shifting.. It's a necessity but things are getting very complicated and somewhat out of control at this stage, especially when you look towards the top end.

A lot of indies and even bigger outfits not willing to trade blows with massive budget endeavours have come in on themselves.. Which is a good thing because the focus is now on fun and gameplay, I played Divinity Original Sin 2 and it was a really good game even though in terms of modern games it's relatively small. Do it right though and there's tons of hours of decent gameplay without grandiose universes.

Whilst procedural creation is become somewhat a necessity in higher tier games (even from lone wolfs) we know when overused results it results in games like NMS which isn't desired, or they get stuck in development hell like Star Citizen..

End of the day, procedural methodologies are just tools.. They can aid you but the new focus on quality over quantity is a good thing, tons of "stuff" and large for the sake of it seems great in theory but in practice it's another matter.

 

What points? I made no claims.

 

The number of assets isn't what is being measured, though, only the size of those assets. Going from 2k to 4k to 8k textures is an exponential increase in file size, but if all of those files are being generated by Substance Designer, then there is no significant increase in labor. It's the same story for lightmaps and audio. How much data they actually take up is entirely arbitrary. Audio in particular has always been compressing it's file size down to whatever point will fit it all on the disk.

 

I'm not sure I necessarily agree on (all) the forecasts. I think the smart AAAs (aka almost all of them) are going to find ways to cut costs. You can see EA trying to do that by implementing Frostbite across all their games. Additionally, Ubisoft uses procedural generation quite a bit in their development (they used it for AC Unity I know, to build the city blocks, and anyone who's seen the video for Beyond Good and Evil 2 knows they have to be using there (though I don't recall Ancel's statements in that video)).

Along similar lines to that second point, an increase in proceduralism can cut costs. My mind immediately goes to procedural textures, which I know nothing about, but seems to have the potential to help with this content creation problem.

(after reading the actual article, he apparently addresses this. Good on him)

Is he really predicting 1 TB games in 2020? That's pretty ridiculous.


Some other things I see from reading the article:

He's including free-to-play games with regular games. That seems like a mistake to me, mainly because such games are designed differently. This is just my opinion, I have no strong logical point to back it up.

He claims we're close to market saturation, and that "we reached 50% of people are gamers long ago." Well, not really. Not really at all, actually, unless by "gamers" you're referring to mobile games. And that's once again an issue like my previous paragraph, of conflating totally different types of game developments as one.

As "proof" let's consider the sales of the breakout "hardcore" game of 2017, PUBG (and regardless of whether or not one thinks it's "easy" or "based on luck," a first person shooter designed around competition between 100 people is absolutely a "hardcore" game). We'll go with this (the most recent data I saw at a glance) which says the game has sold 30 million. Let me first stop and say wow, that's an incredible number.

Anyway, taking a look at this (it's a question on Quora, but the answers link to legitimate sources including the UN) gives us a few estimates on how many people live in developed countries. I'll ignore the fact that the market for games will get larger as underdeveloped countries get developed, and focus on the UN data which claimed in 2010 that 1 billion people live in developed countries.

30 million. 1 billion. 3%. Where in the world is he getting 50% from anyway? He doesn't source that claim, which brings the validity of that and other claims into question, but I found this which looks weak in some areas. As one example, for any landline answers they specifically asked to speak to the youngest person in the house, which seems incredibly sketchy to me. This understandably skews their data towards the younger people, which according to their results has larger percentages of people who play games. The 18-29 age group is 67%, 30-49 = 58%, 50-64 = 40%, and 66+ = 40% (how in the world did they come up with those group sizes?).

Along similar lines, the video games with the highest budgets are also "hardcore" games like GTA V, Destiny, TOR, DICE games, COD games, etc. The "hardcore" market is nowhere near saturated. That's a completely ridiculous assumption.

He makes a good point about marketing costs increasing when trying to promote outside of your core audience. However, a few things like the decreasing "stigma" of games (which is still going on for hardcore games) mean that the core audience is necessarily expanding, so that's a mixed bag.



One of the most concerning things is where he gets his data on game costs, which I don't see anywhere. I hope he's not using VGCharts, which is well known for being inaccurate. If I don't know where his costs are from, can I believe him?

Along those same lines, I see in his first plot has a 360 million dollar game from 2014. The problem is, there are no 360 million dollar games. He's probably talking about Destiny, which we all know was rumored to have a 500 million dollar budget. However, if you look at the Wikipedia page for Destiny, there are comments from both devs and publishers saying that it was nowhere near 500. In fact, the 500 includes multiple things unrelated to the cost of the game itself which could be spread out over the entire IP. According the actual contract for the game (I didn't read through it, I'm trusting WIkipedia which links it) the cost was 140.

The most expensive game listed in this guy's data is completely wrong. Why should I believe him?

On game sizes. This is also a very questionable claim, given what we know about things like compression and high res textures--additional "byte" size that is already part of game development but was typically in times past curtailed at the cost of performance. In more recent times these things have instead been pushed to the forefront to allow for higher-quality performance at the cost of game size. One thing that immediately comes to mind is audio compression--the audio itself (and the work to make it) is unchanged, but with less compression the size goes up.

Let's take a look at a few AAA series over time.

AC Black Flag - 29 GB
AC Unity - 41 GB
AC Syndicate - 40 GB
AC Origins - 43 GB

As one can see, the install size seems to be leveling off.

COD Advanced Warfare - 45 GB
COD Black Ops 3 - 45 GB
COD Infinite Warfare ~ 54 GB
COD WW2 - 45 GB

So once again, there's none of this crazy escalation as implied by his data. The only company I can think of making these truly crazy larger and larger sized games is Microsoft, and I have no idea what's going on with them.

Another thing is his trendlines, which clearly do not represent the data properly in a few cases (the main one being bytes). A linear regression is obviously not accurate there, so I have no idea why he's using one.

I could go on but I'm getting tired of writing this. Basically, this guy doesn't source his data, his data's wrong in some cases, and the conclusions are based off of simplistic statistical analysis. After going through it more deeply I'm less confident (aka not confident at all) in his statements.

 

Take anything Koster posts on his blog with a grain of salt.

His claim to fame was Star Wars Galaxies: An Empire Divided before that game he was relatively an unknown man in the industry. Kind of like how we know who Sean Murray is now, because of No Man's Sky and we will continue too hear about him for a few more years too come, even though he too was an unknown face in the industry before that game was announced and show cased to the public.

 

Indeed. He was a pretty well known at that time, and he was our team’s cd for a couple of years right after that. He’s a nice guy, and knowledgeable, but like a lot of designers that get known for a hit, they have a hard time repeating their success. Mostly now he does articles and speaking about observations of the industry. While interesting, they don’t really provide any practical information for developers, or much that really isn’t already known. Fun to read sometimes for sure, but really just editorial commentary.

 

Battlefield
1942 (2002) 1.2 GB
2142 (2006) 2.2 GB
Bad Company 2 (2010) 15 GB
3 (2012) 20 GB
4 (2013) 30 GB
Hardline (2015) 60GB
1 (2016) 50 GB

* Note this is probably just the initial game install spec and not the DLC additional space needs.

And as technology improves e.g. display resolution, gpu, cpu and sdd/hdd the industry will push to take advantage of the capabilities this inherently pushes up install sizes.

I would expect that there is an even more pronounced install size change in the mobile space...

iPhone (Yr) Ram HD
1 (2007) 128 MB 4-16 GB
3GS (2008) 256MB 8-32 GB
4 (2009) 512MB 8-64 GB
5 (2012) 1GB 8-64 GB
6 (2014) 1-2GB 16-128 GB
7 (2016) 2-3GB 32-256 GB
8 (2017) 2-3GB 64-256 GB
X (2017) 3GB 64-256 GB

Or is your argument based on price or development costs?

 

1. You're comparing games across an enormous time range, and across multiple console "generations." Stick with one. I went with the most recent, which sees games leveling out around 50 GB. Your data supports this.

I agree that size will go up over time. However, there's nothing to support this statement:

"We’re talking one terabyte games that cost $250m to develop, by the early 2020s."

That's utter nonsense.

2. Why are you posting the size of cell phone storage? Both of our examples are based on console/PC games. For both consoles and PC, storage space is significantly larger than our "max" of 50 GB.


My argument is that most of the data doesn't support what he's trying to say.

 

So game sizes increase due to technology but the cost of developing more content will not, what technology offsets the larger workload for developers/artists/modellers.

More data on the price of games over time adjusted for inflation would be ideal, surely the industry has this data?

 

No one's saying the cost of developing the content straight up does not increase with game size increase. But it's not a linear correlation as implied by the graphs. Murgilod and myself both gave clear examples of situations where the development time or "cost" is the same yet can encompass a very broad range of final sizes--texture and audio.

I'm sure there's data on the price. Not sure where it is though, and keep the growth of the market in mind as well--which is probably a very nebulous thing to try to define, so it would be difficult to say whether or not the fiscal return on games (absent post-release monetization like MTX) has increased or decreased.

 

Everyone is an unknown until they do something that gets lots of attention. That's how fame works. I don't see how it's evidence against anyone.

I definitely agree with taking this stuff with a grain of salt, but that's because it's a prediction of the future, not because of how the fellow making it became known.

 

This. Tech and shifting markets are prone to disruption. Trends change constantly.

But more importantly, this is all just high level observational stuff. Some of it is just common sense, just with a pretty graph. But, let’s say these observations are 100% accurate and come to fruition. So... what? How does it impact your development today? Knowledge of broad trends (that aren’t market related) won’t help make a S***ty game good, or a great game fail. It’s not anything that actual studios can have an impact on (or want to). It is completely irrelevant to hobbyists. And if there is some developer out there who is going to be the next big game changer, they aren’t going to be making choices based on theoretical trends, they are going to be breaking the mold via creativity and design, not memory footprint.

——

So, please, let’s not have any arguments or heated debates over one persons speculation of the future that doesn’t impact or can be affected the games we are working on today. It’s interesting information, but nothing productive to argue about.

 

Plus, I suspect that a at least some of it is coincidental rather than either a cause or an effect anyway.

For instance, the thing about dollars per megabyte is surely related to decreases in hard drive manufacturing costs over and above anything else.

 

So it's nothing to do with the rising power of the GPU to display the data?

 

No, because there are a billion other factors that can effect a game's size that have nothing to do with the GPU.

 

I find the prediction about 1TB games in 2020 to be extremely unlikely. A 1TB game would take forever to download for most users, and many users aren't sitting on enough disk space to justify giving 1TB to each new AAA game. Additionally, there has been a large trend toward smaller ultra fast SSD drives instead of slow large capacity mechanical drives.

On the other hand, if he is talking about using extensive procedural systems to deliver the equivalent to what would normally take up 1TB in size, then I do actually think that could be possible. For example, if a 50GB game used procedural systems to massive extend the range of that content to feel like it was actually 1TB of content, then that might be possible.

 

Technically we probably already have that.

 

Wasn't NMS like 6GB?


250 games.
Not complaining or anything but that seems like a tiny sample for 32 years.

 

Yes, but that's the data used to generate the content. If you were to generate every possible combination for an asset and then store them the result would far exceed 6GB.

For another example look at Borderlands 2. There are approximately 17.75 million possible combinations for the weapons in the game. Consider how much space it would take to store every single one of them in generated form.

http://www.eurogamer.net/articles/2012-07-16-how-many-weapons-are-in-borderlands-2

Edit: It just occurred to me you might have been correcting the post about NMS being an example of 50GB...

 

Did "QUOTE" tags give it away?

 

yeah I should have specified more clearly which part I was referring to. Though that only makes it a more notable example.

 

Well those repeated texture and asset reuse are technically lowering the Gb count, could be worse (stare at Id's rage's megatexture, unique details everywhere)

 

For all the talk of NMS it's a shame it fell flat.. It doesn't seem like something one should strive for.

 

Yeah definitely not implying it was a good example of how to do proc gen. Though they've done a ton post-launch to improve things (and I think there's an ARG going on right now with more).

 

No Man's Sky would have been better received if it had been marketed as an early access game. With the way it has continued to receive content expansions people who normally pay for games of that nature would be happy with it.

There is. I received an Atlas Pass in the mail a few days ago.

 

NMS got 24 000 concurrents player after it's biggest update (6 month later), it also have consistently around 1000 players actively engaged at all time and the current ARG they are making is making the community exciting, especially after the name drop by black mirror's writer.

It's worth noting that prior to sony meddling, they were expecting peak 17 000 player at launch, it's was niche from the get go, for people like me who understand the game and the prior legacy of similar game.

Now if we want to toss the big procedural bone I give you Horizon: Zero dawn or spelunky

 

Yeah, if NMS had launched as Early Access with a price in line with that state, then it would have been well received.

 

Horizon ZD in context to W3 was small and the story line was linear (not as many side turns), it was nothing more to me than another AAA title.. Which IMO is a good thing, it's limited format helped it keep context and stopped it from getting boring..

Every AAA outfit uses procedural content pipelines in some sort of way (for major PC / console titles) and has done for many years (like we're talking mid to late 90's), it's only a relatively new thing in indie land.. Even so there's zero chance of competing with a team of > 100.. So as much fun as all this is ultimatley it's pointless.

 

Why do you bring pcg to bigger world, the real use is to cut cost not to make bigger world, the latter is a side effect of efficiency bring by the tools. I'm saying the tools will get more mature over time and do more stuff. Pcg is basically tools, current hot topic pcg is basically pcg level design, aka placement and circulation, with placement already winning (tree and grass being the most visible part of that) and circulation still in infancy (aka pcg lay out tend to be bad). It's been a long time animation is half hand labour vs pcg, with blending being the prior to true pcg solution like Ik rig.

Now the problem of big empty world isn't a pcg problem, it's a game design problem, many game just spread typical gameplay zone with empty travelling area and call it "open world". Big open world aren't new either (elite predate nomansky), but back in the days there was no distinction between travelling and gameplay. People doing pcg focus on scale not design too much so far, but it's not inherent to pcg, it's just people have to realize what the true focus is.

The real obstacle of pcg is basically knowledge and craft.

Square enix just announced they will invest in procedural generation and AI anyway.

http://comicbook.com/gaming/2018/01/24/final-fantasy-xv-director-ai-design-square-enix/

 

Procedural tools are still in their infancy. Almost all of them are still monolithic linear designs with a single output. It's when you start abstracting stuff out that you find new and interesting ways to put those abstractions back together. I have seen very little of that yet in procedural approaches generally. I'm sure there are some working on it, but not out in public yet.

If our procedural tools were up to speed to current technology, content production would be an entirely different world right now. It's just that there aren't many other industries that really need what we need, it's very much a niche. So we don't benefit from the work of others like we do in other areas.

 

Substance is getting pretty mature in the texture generation scale, some very realistic texture were done for destiny 2, indistinct from actual capture.

 

You should check out Houdini if you haven't already, I was sold after the pipe demo.. Using physics to manipulate pipe overlays, extension points for extruding / extracting surfaces.. The one I saw for UE4 also had a level editor system in it to be used as part of the procedural design.

Once you have the base meshes set up you could easily built a city in a day or two, not saying it doesn't require a lot of know how but that's the future.

 

I think he was famous because of Ultima Online.

If anything, SWGalaxies made him more like... infamous/notorious.

 

More context

 

Did not post that his fame was against him; only pointing out his presence in the industry was catapulted forward because of Star Wars Galaxies, and this seems too some how have given his blog posts some Nostradamus like power or influence to many gamers.

When he posts, many gamers believe.

 

Too be honest only Ultima Online fans may have known whom he was before Star Wars Galaxies.

After Star Wars Galaxies was announced and publically shown you can find a massive number of gamers who knew whom he was (and none gamers too, they were just fans of Star Wars).

I did not even know he worked on Ultima Online until well after I was done with Star Wars Galaxies back in 2005.

Mostly because I never played Ultima Online.

 

I never played Ultima Online either and I knew who he was. Soooooo...

 

I played Ultima Online loads and had no idea he had anything to do with it...

 

Cool, now can we agree that our anecdotal evidence does not matter?

 

Game size trends are more likely a function of available disk space and available download bandwidth continually increasing, rather than any comment on the complexity or effort that went into producing the game.

 

Game size is determined more directly to the quantity/quality of the assets used in the game than on what the current HDD tech is or the available bandwidth is for a consumer.

I simply do not see developers releasing a game in 2025 that requires 750 GB of storage, let alone one requiring 1 TB.

 

Yeah more space just means devs can get even more careless with lack of compression.
You don't NEED 8kx8k cup texture that a player will only see once, but who will complain about that when game is already 130GB.

 

If Bethesda continues their practice of poorly optimized texture packs... they might get close. Fallout 4 is 100 GB.

 

I questioned that in my mind, then went to check, and Lord have mercy you're right. All DLC including the high-res textures, 93 GB. That's absurd.

Hard to believe this is the company that gave us Skyrim for 5...