Netcode refers to the networking rules inside a game engine that keep every player’s view of a match in agreement, frame by frame, and that rule set decides whether a match feels sharp or broken. When the system is tuned well, headshots land, combos connect, and movement tracks your inputs in real time. When it stumbles, the same match turns into teleporters, missed hits, and rounds that feel decided before they start. Architecture choices like rollback versus delay-based, server tick rate, and how aggressively the engine forgives latency are the single biggest reason a title feels competitive or unplayable across the public internet.
This guide covers how netcode shapes competitive online play, walking through rollback versus delay-based design, tick rate choices, and why genres like fighters and shooters demand opposite architectures.
Netcode Is the Invisible Layer Behind Every Online Match
Every online match you join runs on a set of networking rules built into the game engine, and that rule set is what the community calls netcode. It governs how your button press becomes a packet, how that packet reaches other players, and how the game reconciles everyone into a single shared reality. Without it, two players in the same lobby would essentially be playing two different games at once, each with their own physics and their own truth about where bullets land.
What Netcode Actually Does
Three jobs sit at the heart of every netcode system. First, it transmits inputs, which means your stick movement, your aim, and your fire button all become data that travels across the network to opponents and to the server. Second, it synchronizes positions, so every client agrees on where your character stands, which way they’re facing, and what state the world is in at a given moment. Third, it resolves hit detection, which decides whether the shot you saw on your screen counts as a hit on the other player’s screen.
Netcode is not the network itself. Your router, your ping, and your ISP all sit downstream of the game’s netcode. The engine decides how to interpret the data once it arrives.
When that decision-making is sharp, lag feels invisible. When it isn’t, the same physical connection that worked fine in one title can feel completely broken in another, and that gap is where most online frustration lives.
That disconnect traces back to how information actually travels, because the gaps players feel start at the packet level.
How Data Moves Between Players and Servers During Gameplay
A round trip from your controller to the server and back follows a tight four-step pipeline. Understanding each step is the fastest way to see why some games feel responsive and others feel like you’re playing through wet cement.
The Input-to-Screen Pipeline
- Input capture: Your controller or keyboard registers a press, and the client snaps that action to the current tick of the game loop, usually 1/60th or 1/128th of a second.
- Packet transit: The action ships as a small UDP packet across your local network, through your ISP, across the backbone, and toward either a dedicated server or another player’s machine.
- Authority check: The server (or the peer acting as host) validates the action against the current world state, applies hit detection, and decides what’s true for this frame.
- State broadcast: The authoritative result streams back to every client, and your screen updates to match the consensus, ideally within the same frame you pressed the button.
Two mechanisms make that pipeline feel instant even when the network is slow. Client-side prediction lets your character move the moment you press forward, without waiting for server approval, and server-side reconciliation is the correction step that snaps your predicted position back to the authoritative truth if you guessed wrong. Together, they hide most of the round-trip cost, and the games that tune them well are the ones that feel tight on a 60 ms connection.
Tick Rate Sets the Ceiling on Responsiveness
Tick rate is how often the server refreshes the world state and ships a new authoritative snapshot. A 64 Hz server updates 64 times per second, which most players consider the bare minimum for modern shooters, and 128 Hz feels noticeably tighter, especially in fast duels where a single frame decides the engagement. Riot’s Valorant runs on a 128 Hz tick server, which is one of the main reasons it has built a reputation for crisp, trustworthy hit registration even on shaky Wi-Fi. By contrast, a game running on 20 Hz or 30 Hz will feel chunky no matter how good your own connection is, because the server simply isn’t producing new truth often enough to keep up with fast action.
Rollback and Delay-Based Netcode Represent Two Opposing Philosophies
Fighting games sit at the sharpest edge of netcode design because one extra frame of delay can break a combo. That pressure pushed the genre to develop two very different answers to the same problem, and the table below lays out the core trade-off between them.
| Feature | Rollback Netcode | Delay-Based Netcode |
|---|---|---|
| Core idea | Predict inputs locally, then rewind and correct when the prediction was wrong | Wait for the network to confirm every input before showing it on screen |
| Input delay on a clean connection | Near zero, matches offline play | Often 3 to 8 frames added to mask latency |
| Visual side effects | Brief, local “rewind” flicker around the corrected character | Noticeable stutter and frozen animations during packet spikes |
| Best connection range | Playable up to roughly 80 to 100 ms one-way | Starts to degrade around 40 to 60 ms one-way |
| Used by | Street Fighter 6, Guilty Gear Strive, Mortal Kombat 1, Skullgirls | Older Capcom titles, most pre-2018 fighting games on console |
Rollback predicts opponent inputs locally so your game keeps running at full speed, and then rewinds the game state to the correct frame once the real input arrives, which is why the visual artifact is usually a tiny flicker rather than a full pause. Delay-based takes the safer route of buffering inputs until the network confirms them, which produces visible stutter but never desyncs. Capcom’s delay-based implementation in Street Fighter V became a rallying point for the competitive scene, and the backlash is part of what pushed the entire genre toward rollback as the default in Street Fighter 6 and most modern releases.
Those trade-offs became decisive once genre conventions hardened, since shooters and fighters impose wildly different demands on timing and precision.
Why Shooters and Fighting Games Use Fundamentally Different Architectures
The split between genres is not an accident. It comes from a different definition of “fair” and a different tolerance for visual artifacts, and the architecture each genre picks reflects that.
Authoritative Servers Versus Peer-to-Peer
Most competitive shooters, including Overwatch 2, Valorant, and Counter-Strike 2, run on authoritative dedicated servers that every player connects to directly. That setup gives the server final say on every bullet, every position, and every damage number, which is essential when a single hit determines a round. The cost is a slightly higher floor on input delay, because every action waits for a round trip to the server, but the trade is worth it for trustworthy hit registration. Fighting games, by contrast, lean on peer-to-peer rollback, where each client must agree on exact frame data for combos to land, and the GGPO middleware library is what made that model practical for indie studios after it proved itself in arcade-era titles.
Latency Thresholds That Actually Matter
Under roughly 50 ms of one-way latency, most modern netcode feels nearly identical to offline play. Between 50 and 80 ms, rollback still hides the lag well, but the rewind windows start to show occasional visual flicker on long combos. Above roughly 80 ms, fighting games become genuinely difficult to play at a competitive level, no matter how good the implementation, because combos and reactions both rely on frame-perfect inputs. Shooters tolerate higher latency more gracefully, since lag compensation can rewind a hit scan a few hundred milliseconds and still produce a believable result.
The Symptoms Players Feel When Netcode Is Poor
Knowing the names of these problems turns a vague “the game feels off” complaint into a specific diagnosis. Each of these symptoms traces back to a different netcode decision, and each one points to a different fix.
- Rubberbanding: Your character snaps back to a previous position after the server overrules the path you predicted, usually a sign that client-side prediction and server-side reconciliation are out of tune.
- Hit-reg misses: Shots land on the target’s screen but fail to register as damage, often because the lag compensation window is too narrow or the tick rate is too low to capture fast movement.
- Teleporting opponents: Enemies appear to jump across the map between frames, almost always caused by aggressive lag compensation reconciling positions across a high-ping connection.
- Input delay: Movement feels sluggish and combos feel unresponsive, which on a strong connection usually points to an over-large interpolation buffer in the netcode itself.
Most of these symptoms get blamed on the player’s own internet, but that’s often wrong. A 30 ms fiber connection can rubberband constantly in a poorly tuned shooter, while a 90 ms cable connection can feel near-perfect in a game with proper rollback. The difference lives in the engine, not the Ethernet.
What Developers Optimize and What Players Can Actually Control
The split between developer decisions and player decisions is wider than most people realize. Knowing which is which stops you from spending money on a new router when the actual fix lives in a setting or a different game.
Where the Engine Tuning Happens
Developers choose the tick rate the server runs at, often trading CPU cost for responsiveness, and they set the size of the packet interpolation buffer, which is the number of frames the client holds back to smooth network jitter. They also set the lag compensation window, which is how far back the server will rewind a player’s position to register a hit fairly. A wider window feels more generous to high-ping players, but it makes low-ping players feel like they’re getting shot behind cover, and that trade-off is the most debated number in shooter netcode design. Patch notes from Valorant, Overwatch 2, and Apex Legends have all called out changes to exactly these values.
What You Can Actually Do
- Use a wired Ethernet connection: Wi-Fi adds jitter that no amount of netcode tuning can fully erase, and a single $10 cable eliminates the most common source of rubberbanding on a home network.
- Close bandwidth-heavy background apps: Cloud syncs, streaming uploads, and game updates running on the same network will spike packet loss at the worst possible moment.
- Pick the closest regional server: A 40 ms connection to a good server beats a 110 ms connection to a popular one, every time.
- Check tick rate before you buy: A game with peer-to-peer rollback or a 64 Hz+ server tick is the single biggest pre-purchase netcode signal you can use.
For diagnosis, an in-game network overlay showing ping, jitter, and packet loss separates your connection from the game’s netcode, and tools like Discord’s voice activity graph can confirm whether the problem is local or upstream. When the numbers are clean and the game still feels off, the engine is the culprit, and no router will save you.
Which leaves a practical question worth answering before the conclusion: where does a player’s responsibility actually end?
The Bottom Line
Netcode is the reason two games on the same internet connection can feel like completely different experiences, and the seven factors covered here, from the input pipeline through tick rate, rollback versus delay-based, authoritative servers, lag compensation, and the player’s own setup, all stack on top of each other. Get the engine right, and even modest connections feel competitive. Get it wrong, and the best fiber line money can buy will still feel like a coin flip. Your sharpest tool is simply knowing which symptoms come from your network, which come from the game, and which come from the architecture in between.
FAQ
What is netcode and why does it matter in online games?
That set of networking rules inside a game engine that keeps every player’s view of the match in agreement. It matters because it decides whether your inputs feel instant, whether hits register fairly, and whether high-ping players can compete without warping the experience for everyone else. Strong netcode hides the network; weak netcode exposes it.
How does rollback netcode reduce lag in fighting games?
Rollback predicts opponent inputs locally so your game keeps running at full speed, then rewinds the state by a few frames once the real input arrives and corrects the result. The visual artifact is usually a tiny flicker rather than a pause, which is why rollback is the dominant approach in modern fighters like Street Fighter 6 and Guilty Gear Strive.
What is the difference between delay based and rollback netcode?
Delay-based netcode buffers every input until the network confirms it, which adds 3 to 8 frames of delay but produces smooth visuals with no rewinds. Rollback runs inputs immediately and corrects later, which feels snappier but shows brief visual flicker around the corrected frames. Rollback is now standard in competitive fighters, while delay-based is mostly legacy.
Does netcode affect input lag and hit registration?
Yes, on both counts. A game’s interpolation buffer directly controls input delay, and its lag compensation window directly controls whether a shot lands as a hit or disappears into the void. A 128 Hz server with a tuned compensation window, like Valorant’s, will register hits that a 20 Hz server with a wide window simply cannot.
How can players tell if a game has good or bad netcode?
Watch for rubberbanding, hit-reg misses, teleporting opponents, and input lag that doesn’t match your ping. If the symptoms appear on a clean, low-jitter connection, the engine is the problem. Community feedback, patch notes mentioning tick rate changes, and verified pre-release netcode impressions are the best pre-purchase signals.
Which online games currently use rollback netcode?
Street Fighter 6, Guilty Gear Strive, Mortal Kombat 1, Skullgirls, Them’s Fightin’ Herds, and the upcoming Tekken 8 all run on rollback. Indie titles have used the GGPO middleware since the early 2010s, and the model now dominates the fighting game genre end to end.
