Reply To: Star Citizen – Musings
CITIZENCON 2018 FULL KEYNOTE PRESENTATION
CITIZENCON 2018
Well amid much fanfare, the keynote of the show which returned to Austin this year, was off to a great start when it crashed after the player interacted with a coffee mug. But no biggie though; after all it’s only pre-Alpha after seven years of development.
Having over-hyped the upcoming 3.3 build which was split into two parts (see my article on that) back in September, all focus was on the neutered 3.3 build which they had said would be released during the show. I am not going to spend 45 minutes writing about how it was basically just another slog through a badly broken tech demo; or how they truly had nothing substantial to show for a whole year of work.
At this moment in time, everything about the game is either incomplete, flat-out broken, or just awful due to bad design choices. The live play through of what we believe to be 3.3.5 (since it contained Loreville which isn’t in 3.3) had the usual great looking set pieces. But broken AI pathfinding, awful character models, assets popping in and out of view, chuggy performance etc, all served to completely negate whatever pleasantries are in the presentation. Having raised – and spent – almost $196 million (as of this writing) on this project, one thing is certain, NOTHING in this presentation shows the quality you would come to expect from even a $10 million game project. Go look at upcoming sci-fi games such Cyperpunk 2077, Rebel Galaxy Outlaw etc and you be the judge.
And there was a train station – with a schedule. The designers really think that making players wait five or more minutes for a train is a great design choice in a video game. But the train was a great way to show off the insurmountable technical limitations that they are facing. But more on that later. After the train ride, we end up at another part of the scene where after getting a mission to recover parts from a crashed satellite, he grabbed a fighter to head to the crash site. Upon arrival at the crash site, and after much screwing around, he dies trying to do a jump from one place to another. A quick dev warp cheat later, and he was back at the site. Shortly after there is a very uninspiring planetary battle against the most boring backdrop and incompetent AI and team mates – ever. If you are thinking “…wait a minute! I saw this last year!“, you would be correct. Same gameplay premise, different “monster in the closet” set pieces.
Oh, as if that wasn’t bad enough, a little over a year ago, in yet another instance of scope creep, they announced that they were going to implement middleware from FacewareTech to bring a gimmick to the game which would debut in the upcoming 3.3 patch. Face Over IP (FOIP) uses a webcam to project the player’s face into their in-game character. They tried to demo it during the presentation, and it was a mess that completely fell apart. With the insurmountable networking problems the game already has, I can’t wait to see the reaction to this one when it gets released wide.
You know what? Screw this – I’m not writing about this Godawful gameplay presentation any further. Go watch it for yourself and be the judge. Enjoy!
STAR CITIZEN 2018 GAMEPLAY. $200M WORTH OF UNINSPIRING TRIPE
STATE OF THE ENGINEERING FARCE
What I do want to write about is something that I have written about since 2015. That being the insurmountable technical challenges facing this project since the scope was increased beyond reasonable comprehension. If you recall, a specific limited scope game was pitched back in 2012. Then, somewhere along the line after Chris Roberts figured out how to monetize scope creep, he kept increasing the scope all the way to a $65 million bounty in Nov 2014. The very same month that he had claimed the two games would be released. All of a sudden, he was building an MMO, complete with in-game micro-transactions with real money transaction hooks. And so the era of the JPEG bounty was born. And over the years they’ve continued to increase the scope beyond those stretch goals, and all the way to almost $200 million. And they’re not even 30% complete yet.
To hear the most devout backers tell it, you would think that CIG somehow invented a swath of technologies since 2012. To them things like first person mode inside fully modeled 3D ships is ground-breaking. This despite the fact that Star Wars Galaxies, Angels Fall First, my own Line Of Defense game etc, already did all that. Large game worlds requiring 64-Bit scale precision? Yup, some of us have actually done that. Battlecruiser 3000AD (which turned 22 years this month) had massive worlds back in 1996 – back when no other game envisioned such a scale. Then along came Elite Dangerous which just took that to the nth degree – and just kept going. Not long after, No Man’s Sky came along. Nested physics scenes with independent collision? Yup, been done before. Any game you have ever played which has a vehicle or building that you can enter, regardless of how many rooms or levels, does that. Right now, you can take any Battlefield game, Just Cause, any Call Of Duty game, any GTA game, or any $9.99 game on Steam which has vehicles and buildings, enter the vehicle or aircraft, crash it into a building or flying around the world, get out, do stuff. I mean, the first time they were promised planets, back in 2015, and then barren moons showed up in the 3.0 build back in Dec 2017, they claimed it was revolutionary. This despite the fact that there are many games with both space and planetary scenes/levels in existence. They think this is all revolutionary, and that we are all collectively fools for doubting that CIG could ever do those things.
Except that none of it is revolutionary; and we never claimed that CIG couldn’t make a game of sorts. Everything they’ve done and are doing, has all been done before, done better; and not only cheaper, but also exist in games that have since been completed and shipped. As I have said time and time again, backers have paid for a Golden chest, but are going to end up with a cardboard box – without a lid.
In my previous article regarding the next phase of the shocking (to backers) revelation, I said:
“Yet they were planning on adding what we believe to be high visual fidelity content to a build of a game which simply would be unable to handle it. And yet still, they totally planned to implement both of the core tech and said content – in the same build. And they did that because since 3.0 (a performance nightmare) they’ve upped the fervor with which they were claiming that both (Network Bind Culling + Object Container Streaming) core techs would solve the performance problems in the next Jesus Patch (that being 3.3). And most of the backer sheep have been hanging their hats on that particular bit of nonsense ever since.
I still maintain that it’s simply not going to turn out that way because that’s not how that works. There is no magic bullet or magic tech; especially not when your engine baseline is absolutely incapable of powering anything remotely as large as what they want to build.“
For the longest time they have been promising backers that the performance issues which have plagued the game since 2.0 was released in Dec 2015 would be resolved with the introduction of new (it’s not) technologies such as Object Container Streaming (OCS) and Network Bind Culling (NBC). I have written several articles indicating that neither of these two is going to solve the performance issues because 1) the visual fidelity of the art assets and scenes are way too high, and level of detail meshes can only get you so far 2) the custom engine built from a potpourri of CryEngine and Lumberyard, simply wasn’t up to the task.
So it came as no surprise that, as I wrote last month, they started walking back those incredulous claims.
OCT 12 2018 DEV SCHEDULE UPDATE SHOWING 3.3 SPLIT INTO 3.3.5
When you look at the presentation, notice how the frame rate drops when he is looking out the window. Also notice how the distant (and some close up ones) assets look very low quality – especially when he is flying around. Some have claimed that they are using 3D textures (like the kind you find in skyboxes) in those outlying areas. They are not. It is a fully 3D level built in such a way that you can’t just travel anywhere, land anywhere, enter buildings etc. It’s all a massive set piece designed to give the illusion of expanse. That’s why they have restricted “no fly zones” even within the confines of the city. Backers who think they’re getting GTA V, The Witcher, Assassins Creed etc, are going to be sorely disappointed. Now you see why the train is needed to link various parts of the “level” because it’s not designed to be open world. Don’t get me wrong, there’s nothing wrong with any of this. The issue is one of visual fidelity vs performance. If you are having fun in a game, you’re not going to notice that you’re restricted to a set piece; after all that’s how most games are designed. But that’s not what they have been promising backers for seven years straight.
In the dev roadmap, they provide basic explanations for what exactly these two technological improvements do.
“Object Container Streaming considerably expands the playable game area and increases performance”
OCS is about memory and loading performance. If you have a machine with 16GB of memory, you will barely have memory (DRAM|VRAM) left for anything, let alone a game that relies on high performance and lots of memory. If assets are constantly being loaded from disk, paged in/out of memory etc, there is a performance hit somewhere in the chain. Especially if the client is loading assets which are outside of the player’s cull distance. If you limit this and reduce the memory footprint, the performance gains are not only barely noticeable in a game that’s already a performance hog, but you will eventually hit a wall where you can go no further.
“This work aims to help improve performance in multiplayer by cutting down the number of entities that exist on clients. Entities too far from a player will be removed from the local client, and when the player moves or the server otherwise detects new entities entering the player’s range they will be added to the client. Because clients will then only consider updating entities that are near to them the overall CPU load will be reduced and performance should improve.”
NBC is all about networking and goes hand-in-hand with OCS – at least in terms of what CIG claims they’re doing. The only reason why this will reduce CPU load is because the client then wouldn’t have to load entities it doesn’t care about. But it’s not going to increase networking performance to any reasonable or noticeable degree because, well, it’s not as if they’re loading assets into packets and sending them across the network. All they can do is reduce the size of packets, the frequency at which packets are sent, and the accuracy (UDP vs TCP) and order in which packets are sent and received. And again, there is a point where you hit a wall because compressed or not, there is a limit.
And according to Chris (@ 4:25), due to instability they didn’t even have this enabled in the live stream build which was running 3.3.5 branch build, unlike the current PTU build which is running 3.3 without Loreville all those bits.
LINE OF DEFENSE HAS 13 CONNECTED SCENES
When you’re building an open world game, you not only need the ability to stream and demand load assets, but you also need to be optimizing every step of the way. If you fail to do these basic things from the onset, then think you’re going to be able to get huge performance gains down the road, you’re just setting yourself up to fail. Open world games like PUBG, Day Z etc, all suffer from this specific performance related problem – and there’s no fixing it to everyone’s satisfaction. Star Citizen has always touted high visual fidelity assets, long before it became this bloated mess that’s stuck in development hell. It’s how they have used high fidelity visuals to monetize the game from the onset.
In LoD, I had the option to have a single cohesive world (with both space and planetary regions) similar to my previous massive games in the Battlecruiser/Universal Combat series. Had I done that, we would end up with large and highly detailed planetary scenes which would have huge performance issues down the road. In fact, compare a planetary scene in All Aspect Warfare, Universal Combat CE 2.0, and Universal Combat CE 3.0 to one in Line Of Defense (better yet, watch this live stream recording from earlier this year). And even after breaking up the LoD world into 13 (space, planet, stations, carrier) sections, we still had to keep the visual fidelity in check because as an open world multiplayer game in which the player can go anywhere in first person mode, vehicle, or aircraft, a performance nightmare is always around the corner. And the trade-off between visual fidelity and performance, is a dance you do with the devil – constantly.
Even in splitting the world, we still had to figure out a way to connect them. So we came up with a custom network architecture with server interconnection (which CIG is calling server meshing) nodes. The way it works is that any of the 13 scenes can be loaded independently on the server. And as they are all linked via jump points, dynamic jump pads, and turbo shafts, what we did was connect all of them via IP hopping. So for example a client would load into Heatwave (planet). In fps mode, to go to Lyrius (space), they would need to take an aircraft, fly up to a jump gate on the planet, then select their destination. Once that selection is made, the client is moved from that scene to the target – regardless of which server is running it. And from space, the client wanting to enter Arkangel station or GCV-Starguard carrier (which we also had to split into three parts), would dock their ship with the target; and again an IP hopping hand-off is initiated, putting the client in that scene. To go from inside the station or carrier to the planet below, the client would either use a ship, fly to space, then select a planetary base target via a jump point, or use a special suit that allows them to jump from orbit directly to the planet surface.
Doing it this way not only keeps performance in check, but it also allows us to run a “cluster” of servers, each hosting a number of scenes. And they’re all not only independent, but you can also hop from server to server based on the target scene. It also allows us to control the number of clients in a scene by restricting access depending on how many clients are already there.
This is not something that you build in the middle or toward the end of a development. It is fundamental to the design of the game; and where performance is concerned, it’s something that you have to worry about right from the start. CIG went the other way. But that’s OK though because backers say the real development on the project only started last week.
Right now those who are playing 3.3 in the Public Test Universe would have seen minimal performance gains depending on the server population. For example, if you’re getting 30 fps in 3.3, when you were previously getting 15 fps in 3.2, trust me when I tell you this, that’s short lived and you really didn’t gain anything. Don’t take my word for it though, wait and see what happens between 3.3 and 3.3.5 which has the new planets seen in the CitizenCon presentation. The performance will again go back to probably worse than 3.2 levels due to addition of these new highly detailed planetary scenes because they add more to the game than ever before. I said the same thing long before 3.0 was released in Dec 2017.
If nothing else, the CitizenCon presentation of 3.3.5 should serve as a wake-up call for what’s coming when they do release it. And it will come along with the same promises from 2015 about “on-going performance tweaks; it’s only pre-Alpha; we’ll optimize later etc“. Except, as has been the case these past three years, it will remain a performance nightmare for the foreseeable future because they HAVE ALREADY HIT THE WALL. And they know it. Which I suspect is primarily why they had to split 3.3 into two parts while they continued working on that.
Bottom line, they’re never – ever – going to be able to do instances with thousands of players as has been promised. Right now they can’t even get 16 players in an instance to have a good experience. As I wrote back in this 2017 article on the subject of the “server mesh” nonsense, if they ever do it, all that’s going to entail is the ability to move clients from one instance to another, as that may be the only way to “stitch” together the game’s proposed world in order to give the appearance of a seamless MMO world – with load times. They haven’t left Stanton yet though – so there’s that.
CHRIS ROBERTS ROAD TO RELEASE KEYNOTE
THE FLYING ELEPHANT IN THE ROOM
Now seven years into the development, Chris attempted to define what a “release” is. And as these things go, he failed miserably. For starters, he was highlighting all the different milestones they hit, starting with the hangar module which debut back in August 2013. Then he got to the current 3.3 and 3.3.5. From that point on, it was the usual waffling, baseless claims, meaningless promises etc. None of which answered the ultimate question: When is the game going to be finished and released?
Most of us who have developed games know that there is nothing worse than engineering debt for a project that the producer seemingly has lost control over. As I mentioned in my previous article, we are going into year seven (eight if you count 2011 prototyping year) and they haven’t even built 1% of the game universe promised. With the upcoming 3.3.5, they wouldn’t have even finished Stanton system which is just 1 out of the 106 promised in the finished game. Not only that, you can’t even fly to any other system, let alone to its planets and moons as there is still no interplanetary travel either. Which makes sense since there are no other systems in the game.
Which begs the question. If they have yet to build the complete game world – or even a fraction of it – how then are they going to complete and release Squadron 42 in the short term? Unless it takes place solely in Stanton, or they are planning on using specific level based set pieces for it, the world that the story takes place in, seemingly doesn’t exist. If it does, they haven’t made it public. And if so, since both games take place in the same world, why not? I have no reason to believe that it exists. Which explains why, technology aside, SQ42 has been delayed year after year. Before I forget, I’d like to point out that in this latest event, he again promised the full SQ42 roadmap coming soon. He said the same thing – last year.
All that aside, once again during the event, in addition to the commercial for a new ship added to 3.3, they unveiled yet another ship concept. That being the Kraken – which never existed before in the game lore or design, and which they’re now hyping to sell – at a high premium. As a JPEG. The most astonishing part of this is that the new ship in 3.3, the Valkyrie, is being touted as the first direct to game addition. In other words, they designed and built it, then immediately added it to the game. You know why this is astonishing? Because they still have a huge backlog of ships which currently exist either as JPEG concepts, or in early model form – and still not in the game. But they had already sold those, and got the money for them. So now they introduce a new one – ready to go in the game – for new cash money. That’s basically an engineering Ponzi scheme.
In a post-show interview, Brian Chambers indicated that even though they have a roadmap and know where they’re going, he can’t say when the game will be completed, when backers will get the “best experience possible” etc. He also said that there about 450 (!) people currently working on the project. Also in a recent broadcast, a backer sent in a query in which he claimed that CIG seemed to have a ship creation backlog of about seven (he was being generous) years. The response from the CIG dev was as funny as it was ridiculous. But it’s in line with the same script from Brian Chambers’ interview.
NO SQUADRON 42 THIS YEAR. BUT HERE’S ANOTHER SNAZZY TRAILER
ROBERTS SPACE INDUSTRIES (UK) FINANCIALS
Meanwhile, over in the UK, earlier this month they filed accounts for RSI, but still haven’t filed for the parent company, CIG which owns RSI and the other companies (of which F42-UK is a subsidiary). Going by the fact that F42-UK was literally insolvent in the their 2017 filings (my article breaks it down), I wasn’t expecting that they would file for the parent company until after CitizenCon due to some of the eye-openers we’re expecting in there. They probably won’t file until after the anniversary and holiday sales, or at some point in 2019. They have a history of filing late, while paying fines; so we’ll see how long they delay this one for.
Regardless, it is a criminal offense to not file accounts within 14 days when due. And right now, both the accounts and confirmation statement are overdue.
In the meantime, someone else over on SA has done the break-down of the RSI filings. I will just quote it below, with minor edits and corrections for clarity and accuracy. You can find my running list of companies over here.
PREAMBLE
This company exists to make accounting mistakes. It serves as the resting spot for money coming from the US and then being redistributed in the UK. The distribution changes pretty much every year. Do they distribute to Foundry 42 or to CIG, who then funnel it onto Foundry 42? Not sure which option they chose this year.
The source of funds from the US changes even more regularly and this year it’s obfuscated.
The only real purpose this company has is to make the “accidental” mistake that the UK group derives 100% of its income in the UK, when in fact the group derives 100% of its income from the USA.
As usual it nets out. A couple of things to note however. In 2016 the company made a gross loss and admin expenses were a negative expense which made up for the gross loss. It was something like £139,487 foreign currency gains with £26,477 admin expenses, meaning once netted off, is the £113,010 contribution to net profit as shown. In 2017 they have not disclosed the foreign currency gain/loss. I wonder if that could be because they made a basic mistake last year :
ACCOUNTING ANOMALIES – 2016
Knowing your debits from your credits may be hard for your first week but this should become pretty ingrained as part of the core of double-entry bookkeeping.
It’s pretty simple. A credit in the profit and loss account is a “good” thing, it increases profit. The corresponding debit in your balance sheet is a “good” thing, it will represent an asset. So a debit value in your profit and loss account will be a “bad” thing that reduces your profit and the corresponding credit in your balance sheet is a “bad” thing that will represent a liability.
So if you make a foreign currency exchange gain, it should be apparent that it will be a “good” thing in your profit and loss account that increases your profit. Thus a credit balance.
So here, you can see the wording is correct in the 2015 RSI accounts. In 2016 someone just had to add “(gains)/.” to the line where it says, “Exchange losses”. Instead they added brackets around the word losses and then added in the gains/ without brackets. This is painful to see because of both how basic the error is and how they had to go out of their way to make the error. I included the line from the CIG UK Ltd 2016 (group) accounts just to show it is possible to get this correct.
This has no real numerical impact but this is building on the narrative about the quality of the preparation of the 2016 RSI accounts, which is some foreshadowing.
So this year they just didn’t bother with the fundamental disclosure because…moving on.
This year they made a gross profit of £744,013 which neatly covers their admin expenses of £744,013. It’s kind of a curious comparison because remember, this company has no employees and serves just to funnel money. We’ll come back to this.
The related parties tell us where in the US the money is funneled from. I’ll start by pointing out that in 2016 there were costs recharged in the period from Roberts Space Industries Corporation of £15,132,933. This matches the declared turnover for Roberts Space Industries International Ltd. You do not net off the negative values for CIG Texas LLC and CIG LLC, you should never net off against turnover so they did this correctly. However, I cannot tell what has happened in 2017. Roberts Space Industries Corporation is apparently no longer a source of turnover and neither are the other two companies shown. So there are a few possibilities.
1. Some basic accounting error and they forgot to disclose it properly.
2. Some other unlisted unrelated company is now funding the UK operations.
We know they have not reverted to funding the UK companies directly, as has been done in the past, because we can see the money coming in and going out in the profit and loss.
CREDITORS
This one is curious. Remember how the company suddenly has large admin expenses? Well, there is also a new miscellaneous tax item of similar value. It’s not social security because this company has never had employees. It’s not tax on income aka corporation tax because then it would be corporation tax. It could in theory be VAT but then it would not be an admin expense and thus coincidence. It would also mean that something had legitimately changed in relation to which companies their income were derived from. We’d then be left with unexplained administration costs.
DEBTORS
The implication here is that this company either transferred too much money to the UK companies or transferred money without a valid invoice. Basically they are messing around again and it hurts to think about what legitimate reason could be behind this happening given that this company only exists to transfer money to the UK companies and it somehow got that wrong. Judging by the related parties it also got wrong the part about receiving money from the USA group.
CONCLUSION
Roberts Space Industries International Ltd is my favorite CIG company.
1. It literally has zero reason to exist and is always littered with basic accountancy errors.
2. It also “hides” numerous “accidental” “errors” which actually have pretty big implications.
I cannot help but feel reasons 1 and 2 are related but I can’t quite put my finger on it.
THE FACE OF SCOPE CREEP
CRYTEK v CIG/RSI LAWSUIT UPDATE
On Sept 28th, shortly after CIG filed a new cause of action in the case, CIG filed their opposition. I am not even going to bother going through that one because it’s the usual nonsense that we have to wait and see how the judge rules. My guess is that she’s going to toss (reject) it because as I explained in my previous article, that section 2.4 which is now in dispute, is vital to certain aspects of the Crytek complaint; and so I don’t see how the judge – who pointed it out in her previous MtD ruling – could grant CIG a dismissal on it.
The good news for Crytek is that as the judge had ruled that discovery is to start, both parties filed a joint Rule 26(f) discovery plan. This means that, regardless of the timeline for the judge’s ruling on the new MtD, they have to start discovery. Though CIG is also asking the judge for a bifurcation of the discovery process – and we have no clue why they’d want that. I can’t wait to read the judge’s decision on that one. We’ve seen for almost a year now that CIG has tried to delay discovery for as long as possible. The judge’s ruling on the MtD, as well as their failed attempts to get Crytek to come to the settlement table, put an end to those efforts. But they’re still trying; and this bifurcation request is just another delay tactic.
It’s interesting to note that Crytek isn’t resisting the CIG request to have certain documents filed under seal. We don’t know what those would be, and we don’t know which of those Crytek would want to make public. But when that fight comes, we’ll be sure to read about it in the filings. One thing to remember is that Crytek isn’t fighting a lawsuit for backers; so they have no incentive to take actions which benefit any parties other than themselves. So if backers were expecting to see things like financial accounting, the number of backers etc in the public filing, my guess is that’s probably some of the most important docs that CIG would want to file under seal; and I don’t see why Crytek would fight to make them public since it doesn’t benefit them either way.
Anyway, on Oct 12th, the judge was supposed to hear oral arguments and to rule on the MtD. But then the day before that, she decided that she didn’t want to hear oral arguments. So she vacated the hearing. Now we wait for her written ruling.
FURTHER READING
Previously: Around The Farce