French computer scientist and fighting game wizard WydD explains Street Fighter V’s cyclical latency on the PlayStation 4, while netcode sleuth DigitalHalftones returns with additional videos to further explain the game’s rollback issues.
Editor’s note: Yesterday we peeked inside the Pandora’s Box that is input latency in Street Fighter V, both online and offline. But it seems that was only the beginning of this tricky subject; here’s a follow-up to provide a closer look at some of the lag issues that SFV players — and competitors — are dealing with while we play!
Fighting game player and tech detective DigitalHalftones recently uploaded a YouTube video dissecting Street Fighter V’s rollback netcode, and his findings revealed a shocking discovery: to make up for one-sided rollback, the game’s netcode would introduce lag to another player should their opponent experience even a slight drop in connection. By connecting a PlayStation 4 console and a PC to one network and manually lowering the framerate, DigitalHalftones was able to prove that the opposing connection would lag in turn. This meant that whatever latency was recorded at the end of a round would be introduced in the following round: in other words, SFV outright creates lag to even out the connection between two players. (Original video follows below.)
Fair’s fair in online play.
DigitalHalftones returned with a second, two-minute video (above) in response to comments about crossplay being a potential factor for the lag he recorded. As in the first video, he used the same setup of a PS4 and PC wired to a single internet connection. He began his subsequent experiment by lowering the framerate from 60 fps to 10 fps on his PC, observing that the PC character was skipping on the PS4 side. He then pushed the framerate back up to 60 fps, and while the PC side smoothed out, the PS4 continued to lag, with PC Cammy still skipping across the screen. However, upon the start of the next round, both sides were completely smooth.
DigitalHalftones explained that this event was due to frames dropped at the start of the round, not his connection’s latency between the two systems. “What’s happening is, at the start of the round, when it says ‘Fight,’ the game is looking at the latency, but it’s also looking at the framerate,” he states over the video. “What happens is, when the PS4 lags behind by a couple of frames at the start of the round, the netcode is thinking that’s actual lag. So, say the game is lagging 1 frame, 2 frames, 3 frames? They’re reading that as milliseconds at a time.” With a single frame being equal to 16.6 milliseconds, three frames would be equal to 49.98 ms of artificial lag, causing the opponent’s side to lag just as much in an effort to even out the connection.
“It’s not crossplay that’s the issue,” DigitalHalftones stated. “It’s the system not being able to run or maintain 60 fps, and the netcode. It’s because of the variable input time for the framerate on the PS4.” As is commonly known, PS4s tend to lag when loading — which explains why rounds get worse for players over time when fighting opponents on the console. Through this additional experiment, DigitalHalftones was able to prove that a variable framerate during the start of a round will incite additional rollback that isn’t part of the game.
The ebb and flow of input latency.
The term “variable” was also recently uttered by fighting game scientist WydD, who used it to describe his findings of Street Fighter V’s input delay on the PlayStation 4. So far, WydD has tested the game on PC, the PS4 Pro, and the PS4 Slim, where he discovered the game’s variable input latency. He now breaks down his findings in layman’s terms, explaining the cyclical nature of the delay and how it affects gameplay.
“Basically, the game has some kind of internal timer inside its engine,” he stated. “When you plot the amount of lag frames over time during my measures, you get this weird cycle.” The cycle in question looks like an active heart monitor, with lag cycling from 4 frames up to 7 frames, then back down to 4 frames once more. “Basically, when you play the game on a PS4 offline — like in a tournament setup — your input lag is around 4 frames, then goes to 5 frames after a while, then 6 frames, and finally 7 frames… to finally drop to 4 frames again.”
While the cycle is relatively constant, WydD observed that the exact time of the switches varies from session to session, resulting in literally variable input delay. “The timing of the steps varies between sessions, but the overall shape is the same, and the length of the cycle is always around a minute,” he explained. While he stated that he isn’t one hundred percent confident of his findings, he hasn’t observed this cycle on PC, citing colleagues Noodalls and DisplayLag as others who have found similar results.
The terror of the input lag cycle.
This variable input delay could definitely have negative effects on players as they duke it out on screen, particularly in regards to hit confirms and reaction timing. “It doesn’t interfere a lot in combos due to the input leniency of SFV,” WydD stated. “The game engine tolerates 5 frames of timing for a 1 frame link. If you’re lucky, you are at the beginning of the cycle and you are generally 3 frames faster in terms of reaction.”
While SFV’s leniency might help give players a fighting chance in this regard, instances can occur that tilt the scale towards saltiness. “There can be other weird instances, like someone presses a light punch and the opponent presses medium/heavy punch one frame after, and the medium punch actually hits anyway because you had a 7 frame delay on the light punch,” he said. “That’s never been seen in experiment, but technically, this could happen.”
WydD also created a series of images that visually represent his findings to better describe this phenomenon. Using Ibuki’s sprite from 3rd Strike, WydD pinpointed the exact frame in which delay begins. Once the command was executed, the screen lost color, and stayed gray until Ibuki’s attack animation was completed. This graphic utilizes his methodology as explained in his YouTube video, but he has since stated that he has improved this setup.
Both WydD’s and DigitalHalftone’s findings indicate that Street Fighter V still has a plethora of issues, at least on the PlayStation 4 platform. With frame drops resulting in actual netcode latency — and worse, the variable, cyclical input delay — the console version of the game leaves much to be desired. As seen in Brandon Chia’s loss in the SEA CPT Online 2, the console version’s build (which is also the CPT tournament standard) can definitely affect players’ performance, potentially leading to whiffed normals and botched reaction times — which, at worst, can cost a competitor a win during an important tournament.
Sources: DigitalHalftones (YouTube); WydD (Twitter)