Or, Yet Another Article About Rollback Netcode
I watched the King of Fighters XV first look teaser as it debuted late at night on January 7th, 2021 alongside the rest of Twitter. The trailer only had six characters in it- Kyo, Shun’ei, Benimaru, K’, Leona, and Mai- and was only about a minute long with interspersed moves, supers, and story cutscenes, but visually speaking it was miles ahead of King of Fighters XIV. Bundled into the same press release was an announcement for an Ultimate Edition of KoFXIV packaged with all of the game’s DLC, a PS4 port of the Code Mystics rollback netcode release of King of Fighters 2002 Unlimited Match, and character reveals for Samurai Shodown season 3. The news was genuinely very exciting- a lot of people had been waiting for a KoFXV reveal since Evo 2019- but as I watched the teaser all I could think of was:
“Man, this game looks sick. I hope they announce rollback.”
There’s probably a large swath of the fighting game community that is tired by now of hearing so-called rollback crusaders talking nonstop about fighting game netcode, as if it’s some sort of magical spell. Even Katsuhiro Harada expressed his frustration on Twitter over his player-base nagging him constantly about adding rollback to Tekken 7, mostly because he alleged that T7 already had an in-house rollback solution and that his audience (including other fighting game developers with reputable pedigrees) didn’t know what they were talking about. Still, all that nagging seemed to pay off, since Season 4 of T7 heralded incoming netcode fixes and online quality-of-life improvements that the game had been lacking for four years. If T7 had rollback at launch, it wasn’t working as intended until the S4 patch; after the S4 patch, the connection quality generally improved, but much like with the Street Fighter V: Champion Edition netcode patch in February of last year, it was difficult to tell at first if players were simply feeling placebo effects or what, if anything, changed under the hood. Also like the SFVCE patch, many players were astounded that it took a global pandemic for developers to finally get onto the rollback train and fix long-standing problems in their games.
(In summary, the main issue with T7’s rollback seems to be that there is an inordinate amount of input delay even on high-quality connections, and that this delay is subject to change mid-match as the connection quality fluctuates- something that no other rollback-based titles do in their implementations. Since delay-based netcode models also change their input delay dynamically, there is superficially no difference between T7’s rollback and the netcode of a delay-based title in the sense that your inputs can often still be unreliable.)
I don’t feel obligated to explain or re-explain rollback netcode to new readers, mostly because so many other people have and I don’t want to split hairs in this article. The fact of the matter is, unlike what Harada and some disinterested players seem to believe, people who experience rollback netcode know that it’s not magic- though it feels like it is. Even if some fans don’t know how it works exactly, we have access to layman and technical explanations from experienced developers who are more transparent about it than the likes of the Tekken team. Beyond that, we can also just experience how good it is for ourselves in a cornucopia of other titles.
At the beginning of 2020, I was most looking forward to playing Granblue Fantasy Versus and Under Night In-Birth Exe:Late [cl-r] while I waited for Guilty Gear Strive to debut in the following year. I figured that even if the netcode was bad, I could practice against whatever few stable connections I found in matchmaking and hone the rest at my weeklies. However, since COVID-19 eliminated local events from everyone’s schedule, I was forced to play both games exclusively online, at which point I quickly grew frustrated by the archaic delay-based netcode. This frustration was felt by other players, which meant that the player-base for both games dwindled. When there are fewer players in matchmaking, the game prioritizes finding you matches quickly rather than matches of close skill or high connection quality; this results in you taking forever to match up against someone with a 6-frame delay or higher just to get bodied because your DP or tight combo link didn’t come out when you wanted it to; you drop the game in disgust, and now there are even fewer people in the matchmaking cycle. Eventually, I even dropped my main title, Pokkén Tournament DX, because despite the ongoing developer support for tournament circuits, it wasn’t worth playing a game that had ‘good’ delay-based netcode or was missing basic online features like player rooms and spectate mode when I knew there were better experiences out there. Even while I was putting together Trainer’s School, I saw little value in participating in the game competitively or keeping myself de-rusted.
Because I was unable to play neither GBVS nor UNICLR, I ended up switching over to Them’s Fightin’ Herds for most of the year. While I had owned the game for a while, its 1.0 release debuted early in 2020, so I decided to try it out again now that it was finished. Among other things- intuitive controls and engine mechanics, gorgeous cartoon presentation, fleshed-out singleplayer content, and one of the most robust tutorials I’ve ever played- the rollback netcode alone kept me hooked on playing. I could join a pixel lobby, enter training mode and matchmaking, and then fight whoever approached me. GGPO frame delay settings guaranteed that my input delay would be static and never go higher than 5, and even against the most unstable 200+ ping connections, I rarely had to adjust it to a value higher than 3. This meant that my controls felt tighter and more responsive, and I could practice against a wider variety of players closer to my skill level. As a result of being able to play against more people, I could play for longer and have fun doing so, since I wasn’t getting upset over being matched up with terrible connections or having to wait a long time to play. This also goes for the rest of the player-base who are likewise retained into the game’s ecosystem simply because they know that they can find high-quality matches whenever they want.
Several other games last year also received rollback netcode because of retrofitting, either officially or via fan projects: Super Smash Bros. Melee via the Project Slippi Dolphin injection, KoF02UM via a Code Mystics-developed rollback solution, Fighting EX Layer via an in-house beta and subsequent patch, and Guilty Gear XX Accent Core +R, perhaps the most ambitious success story of all these games. Originally a fan-made GitHub repo in the same vein as the aforementioned Project Slippi, the development team (Team French Caliburst) was quietly contacted by Arc System Works with the purpose of putting them on payroll and making their project official. Within six months of the initial April 2020 trailer, TFC officially announced that their fan project, initially dubbed GGPO+R, would be officially added to the Steam version of +R along with additional features, namely matchmaking updates that brought it up to parity with the 2019 Switch port.
You don’t need to look any further than GGPO+R to see just how important including rollback is to the health of your game. On October 29th, 2020, the day the rollback beta was initially added, the average player count for +R on Steam jumped from 12.8 players to 187.1. The all-time peak players for +R spiked from 428 (from May 2015, when +R first released on Steam) to a whopping 2,330! Throughout the month of November, +R hit its highest average player count of 655.6 players, and even after the initial shine of the rollback beta wore off, +R has been able to maintain an average and peak player count of ~200 to ~500 players, respectively. Speaking for myself, I was playing +R almost every night with friends in Discord, setting up rotations as best as we could in the six-man player rooms, sharing new tech and flashy moves that we were learning with our prior Rev2 experience. It almost felt like we were at our locals again, and the netcode making it so that our opponents may as well have been sitting right beside us certainly helped. My amount of playtime in +R over the course of a few months quickly eclipsed the amount of time I had put into TFH over the course of an entire year. The presence of such a massive influx of new players was even felt in other parts of the online FGC, as the Dustloop wiki was unusable for days because so many users were consulting it as a resource for learning +R. The GGPO+R update brought an entirely new group of people into an eighteen-year-old game, whose most recent update was seven years ago, and was nigh-unplayable online for five years. Rollback saved Guilty Gear.
If 2020 was the year that the FGC and developers learned about how important rollback was for the health of their scenes, 2021 will be the year where developers will see that adding rollback pays off. Guilty Gear Strive is on the horizon, with a slated April 2021 debut and another online beta (date TBD at the time of writing, hopefully featuring the completed UE4 rollback engine) around the corner. Both GGPO+R and the plethora of Code Mystics-updated SNK retro titles will serve as testbeds for their respective companies’ upcoming titles, even if completely different teams are working on the netcode for the modern games. GGPO+R will hopefully be indicative of GGST’s netcode, and KoF02UM’s rollback will likewise be indicative of whatever netcode solution awaits us in KOFXV.
Rollback goes far beyond player retention in new titles, however; it also affects what the greater FGC will be playing in tournaments. Despite a vaccine for COVID-19 entering early distribution, the pandemic is still rampant across the globe, especially in the United States, meaning that the esports ecosystem (no, not that one) will be forced to shift its focus to online events. The FGC is no exception to this. Frosty Faustings XIII, a Guilty Gear-centric Midwest FGC major, has recently announced its plans for an online event, as well as opened signups for its online brackets. The selection for the online bracket is highly indicative of the state of the FGC going forward in 2021, as all but two of the titles available for open registration have some sort of rollback netcode implementation, either natively or via Fightcade: +R, T7, TFH, Mortal Kombat 11 Ultimate, Power Rangers: Battle for the Grid, Fantasy Strike, Skullgirls, Windjammers, Street Fighter 3: Third Strike, and Street Fighter Alpha 2. (The two outliers, Ultimate Marvel vs. Capcom 3 and Tatsunoko vs. Capcom, are being played through optimized Parsec and Dolphin solutions respectively.)
This mix of modern, retro, and indie titles is a demonstration of how much the landscape of the FGC has changed due to offline events no longer being an option and delay-based titles not being suitable for competitive online play, and the lineup will reflect what other online tournaments will look like in 2021. With the cancellation of Evo 2020 Online due to allegations of sexual harassment levied against individuals such as Evo co-founder Mr. Wizard and Lab Zero developer Mike Z, this also means that FFXIII is the first litmus test of how successful a large online major featuring multiple titles can be. (While there have been official circuits such as Capcom Pro Tour Online, large majors for single games such as Grand Stampede, and online local weeklies such as Wednesday Night Fights, FFXIII is a large national grassroots event featuring more than ten titles in its lineup.) We can expect events like Combo Breaker, Final Round, CEO, and whatever future iteration of Evo to follow suit, adding other titles like SFVCE, KoF02UM, Killer Instinct, and so on as they see fit. With high hopes for GGST’s rollback implementation, we will likely see that game added to various online tournaments and circuits after its debut as well.
When a game’s online experience is solid and can replicate offline gameplay seamlessly or as close to it as possible, more people will stick around and play against each other. But it also means that online tournaments for that game carry more legitimacy. If online players can successfully perform all the precision inputs that they practiced offline, the experience they develop in online situations becomes just as valuable as if they were playing at an in-person event. This snowball effect also impacts more than just the players. If online events are more legitimate, then the TOs, streamers, and other staff hosting them can develop their infrastructure and continue to host these events as alternatives once offline events resume. If there are more online events, then there are also more opportunities for casters, journalists, and other content creators to provide commentary and other coverage for these events. When these all combine seamlessly, more eyes are on the scene, including prospective sponsors, and thus more players are watching, which bolsters the entire community even in the absence of offline tournaments.
If a fighting game launches in 2021 or later without rollback netcode or other crucial online features, most players will either play it for a month before dropping it for something else, or ignore it entirely, because they know that other games exist in the genre that will respect their time spent against other opponents. Likewise, if games like GGST, KoFXV, and the in-development UNICLR rollback update end up having successful rollback netcode implementations, the player-base for those games will expand drastically, retaining both newcomers and series veterans. When new players enter the scene, they want to compete, and so online events will provide that avenue; with more eyes on online events, even more people will continue to enter the scene, and so the community grows steadily and organically. The ‘rollback crusade’ is not just a matter of throwing around a buzzword at developers on Twitter until they give in and either block you or add good online to their game- it’s a matter of whether the FGC will continue to grow or will be forced to stagnate in 2021 and onward.
And if it means anything, rollback netcode in KoFXV is a bigger selling point for me than Mai’s boobs. (And I prefer B. Jenet anyway.)