The Star Citizen Thread V10

[Ian Skippy]

Dude, we had smooth movements in online FPS games when playing in on a 386 with a Voodoo 2... or earlier. Why does CI!G get a pass on this when they have been working on it for 6 years and have over 200 million dollars? Especially when the game engine does it flawlessly out of the box?

Voodoo 2 on a 386?

 

[Agony_Aunt]

Voodoo 2 on a 386?

Thinks back

Oh yeah, was probably an early Pentium.

Leave me alone, the old grey matter isn't what it used to be.

 

[Cobra1984]

It would make their life easier if they slowed everything down to realistic speeds to start with. Players moving at human speeds (and even less with heavy armor on..), ships having non-instant thrust availability etc. More Arma, less Quake. Having all that would allow to have both client based movement and less rubber banding due to better server based movement prediction (or interference). There are some very well known algorithms to do that properly.

There are of course industry standards for them to solve all of this.

But the takeaway here is that their state system and animation system were not designed with networking in mind.

Would anyone be interested in a veteran game dev's take on this?

 

[Stigbob]

There are of course industry standards for them to solve all of this.

But the takeaway here is that their state system and animation system were not designed with networking in mind.

Would anyone be interested in a veteran game dev's take on this?

Yep, post the thing or a link.

 

[Navigare Necesse Est]

There are of course industry standards for them to solve all of this.

But the takeaway here is that their state system and animation system were not designed with networking in mind.

Would anyone be interested in a veteran game dev's take on this?

Sure. Always nice. Tynan did a couple on Rimworld. I particularly remember his woes with A* and how little pixels tend to prefer certain paths and patterns. Very interesting.

 

[Cobra1984]

This is the response from an industry professional to a recent post by a CIG staffer regarding their current networking implementation and plans to improve it.

HOLY (removed)

So they've managed to take the absolute worst parts of lockstep deterministic simulation, none of the benefits, and slap it into client-authoritative netcode. AMAZING

First off, this is explicitly a visual delay on remote clients that they're referring to, which is better than lockstep determinism in the sense that the player shouldn't have a local input delay, but worse because we already have the most up to date data for other clients but it's just not being used because they never designed their state and animation systems with networking in mind.

In a lockstep deterministic simulation such as in Age of Empires 2, Supreme Commander, etc., there is a similiar built in delay to all input. This is so that the input can be replicated over the network to everyone else's simulation - so long as no one has a connection latency of > 250ms (or whatever the delay is), it works great.

Since those simulations are deterministic, if someone is cheating and sending bad input, it will desync the game because the cheater's units did 10k damage or whatever when everyone else's client thinks it did the correct amount of 500. The simulation state is hashed and compared each frame to check this.

It's a tradeoff - it's a really good architecture for RTS because then you don't have to sync anything at all except player input, instead of 1000s of unit states, but comes at the cost of that added latency and reducing the simulation's max speed to that of the slowest machine in the session.

Artificially introducing an input delay is probably the worst thing you can do for playability in a twitch based game, and they're not even getting the benefits that come from lockstep determinism.

And to top it all off, it's client-side authoritative in TYOOL2k19, which we already knew because Star Marine was hacked to a hilarious degree within days of its release. Something that wouldn't be possible if they were actually taking advantage of that delay. Instead, they just introduced it as a band-aid solution to their horrible netcode.

All of the problems that they need to solve are described in this documentation:

https://developer.valvesoftware.com/wiki/Source_Multiplayer_Networking

They can be used with OCS or whatever their fancy buzzwords are; there is absolutely zero reason to avoid the industry standard best practices and cook up whatever it is that they're doing, because it's going to be worse.

 

[Cobra1984]

Now if you go back and read what the CIG programmer said, you'll see that what they actually intend to do is to make the band aid (250ms delay) as small as possible (2 frame delay) - instead of actually implementing any of the industry standards and solutions.

THIS is why they are 7 years deep into development with no released games and nothing resembling commercial software. They don't know what they're doing.

All of the excuses posted here and on reddit, that games like these take X years and this is all normal game development suddenly fall apart and fly out the window.

It's an art and animation studio with no clue, no experience and no technology to build the game they've been selling.

 

[Backer42]

CIG made a turn-based FPS.

Spoiler

with 250 ms turn duration

That's an industry first!

 

[Tippis]

what they actually intend to do is to make the band aid (250ms delay) as small as possible (2 frame delay) - instead of actually implementing any of the industry standards and solutions.

To be fair, this is SC we're talking about. Once they've layered in the fidelity, the odds are fair that those two numbers will end up meaning the exact same thing anyway…

 

[Asp Explorer]

Talking of experience - did the latest ATV really say they are hiring people with zero experience and fast tracking them through the CIG school of not making games, to not make games?

 

[Skizomeuh]

Talking of experience - did the latest ATV really say they are hiring people with zero experience and fast tracking them through the CIG school of not making games, to make things never been done before*?

FTFY

*Maybe for a (ton of) reason, but shhhhh...

 

[Cobra1984]

It's ok, they did an internal training day for the new artists.

 

[Agony_Aunt]

Talking of experience - did the latest ATV really say they are hiring people with zero experience and fast tracking them through the CIG school of not making games, to not make games?

No idea because didn't watch, and maybe you just misunderstood, but my impression is, most serious professionals with any knowledge of CI!G wouldn't consider a move to them to be a good career move. On the other hand, someone was reporting that CI!G last year hired some top programming talent... but there again, top talent can afford to go where the money is and move on again afterwards... so anything is possible.

 

[Stigbob]

This is the response from an industry professional to a recent post by a CIG staffer regarding their current networking implementation and plans to improve it.

The best thing about that is I now understand why all my little robots seem a little reluctant when I boss them about in supreme commander.

 

[GuruNot]

Don't forget to by your valentine a new jpeg:

https://robertsspaceindustries.com/comm-link/transmission/16940-Valentines-Day-Promotion-2019

Some of the comments made me chuckle, especially:

Spoiler

Fancy thinking something they just bought would actually be in game

Clearly new to SC.

 

[Cobra1984]

They were supposed to have hired 4(?) top tier programmers in the UK last year, evidently they aren't working on the PU.

 

[AlPilg]

Some of the comments made me chuckle, especially:

Fancy thinking something they just bought would actually be in game

Clearly new to SC.

Yeah, just because they've put something up for sale don't think you'll get anything for your money right now, flippin noob.

 

[MTBFritz]

Just a reminder that there are no sales for Star Citizen...just pledges so whoever thinks he is entitled to anything for all that money clearly has no idea how this ̶s̶̶c̶̶a̶̶m̶ project works.

 

[Asp Explorer]

They were supposed to have hired 4(?) top tier programmers in the UK last year, evidently they aren't working on the PU.

See, that confuses me too.

I cannot remember exactly where I read it, but there was a Genuine Roberts quote along the lines of "We've hired the best network guys for the final polishing stages to ensure the networking is as good as it can get", contrast with the earlier "Network guys are impossible to find", and giggle helplessly at the even earlier (and oft repeated) "1000's of players in the same instance!"

Mind you - even if they hired the best socket slingers on the planet - if all development has focussed on SQ42, what sockets are they going to sling there?

Or perhaps they are kept in the basement, polishing jpegs.

 

[Surefoot]

No idea because didn't watch, and maybe you just misunderstood, but my impression is, most serious professionals with any knowledge of CI!G wouldn't consider a move to them to be a good career move. On the other hand, someone was reporting that CI!G last year hired some top programming talent... but there again, top talent can afford to go where the money is and move on again afterwards... so anything is possible.

That varies a lot, most software programming is done on the lowest skill possible with teams of underpaid juniors (and on offshore locations when possible). Being the lone hero-coder in there only helps them keep the project above the waterline, usually. It's a very, very rare case when the stars align and us veterans are given a blank slate and a team of other like-minded vets to do something really neat. The bulk of software code i see is sub par, and i know video game code is even lower quality than everything else... Which leads to my next thought, the best jobs for high skill coders are usually in fin-techs and other leading industries, it's a well known fact that videogame industry is a low pay coupled with huge overtime hell.