What is ‘rollback netcode’ in fighting game?

If you’re interested in fighting games these days, you’ve no doubt heard the term “recovery netcode” thrown around grapes. Fighting game developers like Arc System Works are making this feature a big selling point in some of their recent fighting games, and even Riot Games has tapped into a restored netcode training program. for his upcoming fighting game, LOTS project.

Fighting game players are always looking for fast and consistent online connections. Since many players can’t go to offline events and play games with other players, online multiplayer is often the only option available to them. For fighting games especially, a good connection means you’ll be able to make your inputs, combine your attacks, and react to enemy mixes.

This means that playing the best possible online is essential to maintaining a healthy player base for a game. Fortunately, in recent years, developments in creating smoother, faster online experiences have blossomed in the form of ‘recovery netcode’. To understand what a recovery netcode is, it’s important to understand the difference between rollback and latency-based netcode.

What is Netcode based on latency?

First, the basis of a fighting game is how attacks are measured. It’s almost like a measure of time in fighting games. For example, one character’s attack may appear for 3 frames, while another character’s slower attacks may appear for 5 frames. When you press a button on your controller, the computer reads it as input and executes the attack. Logically, when two players press a button, the game receives both inputs and processes them at the same time.

This is consistent in offline play and is why fighting game players prefer to practice on the net offline. If you think about it logically, there is always a delay when information is sent over the network. Fighting games need to compensate for this delay so your input doesn’t arrive late and arrive at the same time as the opposing player. (Imagine if you pressed a button and you had to wait a second before your attack appeared, while your opponent was able to escape their attack sooner.)

Latency-based netcode was the formerly popular way to compensate for this delay over the network. Basically, what happens is, the game will artificially delay the inputs for both local and remote players. For example, if a move is supposed to appear in 5 frames, the game can artificially force it to appear in 8 frames. This same restriction is placed on both players. Players who are used to playing offline will feel this difference, which is why netplay is often mocked by the community. (This is a simplification of what actually happens.)

In theory, if the Internet were a consistent battleground where volatility had been muted, this wouldn’t be so bad. Sometimes delay-based netcode runs well and is said to be better than recovery netcode. (This is considering the distance between two players.) However, as we all know, the internet is not a level playing field. Volatility happens all the time.

Imagine if your connection spiked and the information from your controller could not reach the network. You press the A button, but due to your internet spike, this cannot reach the enemy players in time. What happens? The game freezes and slows, because this is the only way to ensure that all inputs arrive at the same time.

What is the recovery netcode?

So what is a recovery netcode, and why do fighting gamers rally behind it as a superior processing power when playing online? Remember the Super Smash Bros. community. How surprised was it? after the developers refused to incorporate the restore?

Basically, the recovery netcode eliminates latency by predict your input. The game will try to predict your input on its own and will show that input to the opposing player. For example, if the game expects you to use Ryu’s Fireball on an enemy, that’s what the game will show the enemy. However, imagine if you used a different attack instead? The game will “go back” and adjust to what really happened. The game rewinds to another state, one where your other attack was actually performed.

So what is the benefit of the recovery netcode? Doesn’t this mean that the game will constantly rewind into different states? In fact, most of the time, this rewind happens in a period of time that is too fast to even notice. If there is a spike in lag and the game correctly predicts all the inputs at that point, you won’t notice a change in gameplay.

In fact, games often predict certain frames in the future with great accuracy. Imagine that you are blocking a long chain of pressure from the opponent. What is the game’s best prediction of what you will do in the future? Will you try and compete with the enemy or will you keep blocking? In most cases you will choose to block and the game will correctly predict that.

When implemented well, restorative netcode can make for a faster, smoother game experience for everyone. Guilty Gear: Strive has a great online netcode and tournaments between the US and Japan have been held to great success. This wouldn’t be possible in netcode based on delay, because the delay would be too strong.

Ars Technica has gone much deeper into what makes recovery netcode so great and delved deeper into the perception of how these two types of netcode work. If you are interested, check out their posts. YouTuber Core-A Gaming also has a helpful video on recovery netcode, including visual examples and interviews with game developers. You can visit his channel or watch the video below: