Sean Tracy · Technical Director of Content

I sort of talked a tiny bit about this during the presentation, and to answer the question directly, yes, Theaters would totally suffer from this. Just the same way that the PU suffers from it, the same way that Squadron suffers from it, the same way that Arena Commander suffers from it, the same way that Star Marine suffers from it. And to go to that point, just first before I actually go into what we've done about it: this isn't a special code base or anything, right? We're not separate from the game. That defeats the point. 100%, right? We don't just develop this separately. We are intrinsically linked to the PU, Squadron 42, and any other game within here. The nice part about that is then when we go to prioritize one of these things like hit registration, or more aptly basically just the lag between the responsiveness of a shot and then somebody dying, it's going to affect everything. So the nice thing about working on something like Theaters is that that becomes front and center very, very quickly. And I would argue that in the PU it's not immediately obvious sometimes. Yes, once you get into those combat scenarios, you absolutely feel that. But just to go to my point about why we need a more analytical environment, that happens within the first 10 seconds of play. And if that happens in the first 10 seconds of play, you're going to want to do something about it immediately.

Jared Huckaby · Host, Community

Theaters will also let us, with these shortened, condensed game experiences, recreate that experience over and over again. It lets us reproduce them, which is the best way to track down the source of any bug, of any performance issue. Reproduction, reproduction over and over until you find the cause. So again, in the PU, those things exist, but they happen a bit more spread out, so it doesn't feel like it's as big of a problem. We condense it, it becomes a greater problem, we can better find the source of those things.

Sean Tracy · Technical Director of Content

And then we know if we've made a difference. That's the other thing, right? And again, sure, in the PU, you could go get four QA guys together against four other QA guys, and they'll meet on Daymar, and they'll have a little combat or whatnot. But again, it's the least analytical way you could do this. We've got to be a little more scientific about this, especially when you're talking about network latency. And you can't just do it with AI either. So a lot of people go, well, just jump on a planet, spawn 60 AI, and start shooting them, and you'll see how unresponsive that is. It's not immediately obvious when you do it that way. What you really need is a full server of players using all our assets. Because, again, the way the grenades work is a little bit different than the way a rifle works, a little bit different than the way a ship works, a little bit different than the way the ground vehicles work. So, having said all that, we actually made some significant improvements. And as I do, I like to take people through the history of the problem up into the actual fix. So about the 2.0 timeframe, we had upwards of a 500 millisecond delay, and this was injected into the system. There were some reasons for this, and it can be argued whether those were great reasons or not great reasons, but it was really about smoothing out actor movement as well as prediction for the AI ships in the verse. So people that were playing in the 2.0 branch, let's say, will remember times when, heck, we even had it in 3.0 one time, where the AI ships were jittering a lot. So again, they would be along a path and it looked like they were rubber banding, right? They would go forward, back, and then forward and back and jitter through. That injection of that 500 millisecond delay was what was meant to actually smooth that out. But it was never meant to be a permanent fix. It was kind of a temporary thing. But like every bad hack, it becomes permanent, right? When you do a quick hack, you'll fix the problem and then you forget about the problem and you move on to something else. So we had this 500 millisecond delay about the 2.0 timeframe. Then in 3.0, one of our gameplay programmers coupled with one of our network engineers, this is Gordon and Jens, reduced this down to 250 milliseconds. So now we're getting a little bit better, but 250 milliseconds comparatively to other games, just to give you an example, popular EA game Battlefield, they're usually running at 100 milliseconds. And this isn't necessarily when you look at your ping, this isn't what we're talking about so much. This is your latency between you, server, client, and then back, right? So there's this thing called rewind time that we've worked on. The first thing we did, though, because 250 milliseconds is what everybody's been playing for the last little while, and just that amount of time is a lot, when you're talking about, again, the immediate feedback of a shot or throwing a grenade or something like this. So we were sitting at the 250 millisecond delay at CitizenCon, and people did experience that, but because we were keeping it to 40 players and not a whole ton of ships, it wasn't super duper obvious, but I think that it was definitely felt. So having said that, the first thing we did when we got back is that the same network engineer actually went to go look at the upstream refactor, and this is actually on the roadmap, and we've been working on this for a little while. So what it does is that it massively benefits a server that is running at a high frame rate now. So what we've done is that he's reworked the way the upstream stuff works, so that's when you're sending up to the server. And when the server's running at a decent frame rate, 30 plus anyways, that latency is going to be much, much less. So you're going to feel really good responsiveness there. However, it won't really help the PU massively yet just because the server frame rates need to go higher. And as server OCS is rolling out, that should be helping that. The next level of this was a huge win, and this is the one that we're really now playing with internally, and this is what's going to make the biggest difference, because now we've reworked the, we call it rewind time, basically, so this prediction kind of stuff, and we've got it down to that 100 milliseconds, but it's not just 100 milliseconds and we try to keep it there, it's actually dynamic now. So what we've done is we've exposed it so that it can actually change, so we can go plus 120, or we can get as long as 50 in some cases, depending on your connection to the server. And then further to that, we can actually tune it now in real time. So as we're playing within the servers, we can actually change this rewind time. So that's what myself and Rich Tyra and Jens and Gordon and everybody are actually looking at pretty much weekly right now: what are our values looking at? What's the range of this? Where should this be? But it's to get it down to the feel that everybody's used to from other games anyways, which is generally around that 50 to 100 milliseconds. And that's where we're getting it to. So basically, when they get hands-on on this, and when it rolls out all together, Arena Commander is going to feel better. Star Marine is going to feel better. PU is going to feel better. Everything is going to be much more responsive. And we've already got that work, not completely done, it's not super tuned, but we've already put all this stuff in. So the next release should actually have a bunch of this in it.

Jared Huckaby · Host, Community

When you say the next release, you mean 3.9?

Sean Tracy · Technical Director of Content

Yeah, I want to be sure about that. I'm going to say 3.9, but it's also sitting in a branch right now, so it's not yet in the main branch. When Theaters goes into their hands, they're going to have it.